1: %\documentclass[useAMS,usenatbib,usegraphicx]{mn2e}

2: \documentclass{emulateapj} 

3: %\documentclass[manuscript]{aastex}

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

5: %\usepackage{graphicx}

6: 

7: % User defined command

8: \newcommand{\msun}{\mbox{M$_{\sun}$}}

9: \newcommand{\msunpctwo}{\mbox{\, M$_{\sun} \, {\rm pc}^{-2}$}}

10: \newcommand{\natd}[2]{\mbox{$#1 \cdot 10^{#2}$}}

11: \newcommand{\pder}[2]{\frac{\partial #1}{\partial #2}}

12: \newcommand{\pdert}[1]{\pder{#1}{t}}

13: \newcommand{\mcorr}{\bf}         % highlight corrections

14: \newcommand{\mcorrm}{\boldmath}  % highlight corrections

15: \newcommand{\bl}[1]{\mbox{\boldmath$ #1 $}}

16: \newcommand{\rhat}{\mbox{\boldmath$ \hat{r}$}} 

17: \newcommand{\phihat}{\mbox{\boldmath$ \hat{\phi}$}} 

18: 

19: \slugcomment{Submitted to Astrophysical Journal}

20: \shorttitle{Disk masses around young stellar objects}

21: \shortauthors{Vorobyov} 

22: 

23: \begin{document}

24: 

25: \title{Disk masses in the embedded and T Tauri phases of stellar evolution}

26: \author{E. I. Vorobyov\altaffilmark{1,}\altaffilmark{2}}

27: \altaffiltext{1}{Institute for Computational Astrophysics, Saint Mary's University,

28: Halifax, B3H 3C3, Canada; vorobyov@ap.smu.ca.} 

29: \altaffiltext{2}{Institute of Physics, South Federal University, Stachki 194, Rostov-on-Don, 

30: 344090, Russia.} 

31: 

32: %\maketitle

33: 

34: \begin{abstract}

35: Motivated by a growing concern that masses of circumstellar disks may

36: have been systematically underestimated by conventional observational methods, 

37: we present a numerical hydrodynamics study of 

38: time-averaged disk masses ($\langle M_{\rm d} \rangle$) around low-mass

39: Class~0, Class~I, and Class~II objects. Mean disk masses ($\overline{M}_{\rm d}$)

40: are then calculated by weighting the time-averaged disk masses according to the 

41: corresponding stellar masses using a power-law weight function with a slope 

42: typical for the Kroupa initial mass function of stars.

43: Two distinct types of disks are considered: self-gravitating disks,

44: in which mass and angular momentum are redistributed exclusively by gravitational torques,

45: and viscous disks, in which both the gravitational and viscous torques are at work.

46: We find that self-gravitating disks have mean masses that are slowly increasing along the sequence

47: of stellar evolution phases. More specifically, Class~0/I/II self-gravitating 

48: disks have mean masses $\overline{M}_{\rm d}=0.09$, $0.10$, and $0.12~M_\odot$, respectively.

49: Viscous disks have similar mean masses ($\overline{M}_{\rm d}=0.10-0.11~M_\odot$) 

50: in the Class 0/I phases but almost a factor of 2 lower mean mass in the Class II phase 

51: ($\overline{M}_{\rm d,CII}=0.06~M_\odot$). In each evolution phase, time-averaged disk masses

52: show a large scatter around the mean value.

53: Our obtained mean disk masses are {\it larger} than

54: those recently derived by Andrews \& Williams and Brown et al., regardless of the 

55: physical mechanisms of mass transport in the disk. 

56: The difference is especially large for Class II disks, 

57: for which we find $\overline{M}_{\rm d,CII}=0.06-0.12~M_\odot$ but Andrews and Williams 

58: report median masses of order $3\times 10^{-3}~M_\odot$. 

59: When Class 0/I/II systems are considered altogether, a least-squares best fit yields 

60: the following relation between the time-averaged disk and

61: stellar masses, $\langle M_{\rm d} \rangle = \left( 0.2\pm 0.05 \right) 

62: \langle M_\ast \rangle^{1.3\pm 0.15}$.  

63: The dependence of $\langle M_{\rm d} \rangle$ on $\langle M_\ast \rangle$ 

64: becomes progressively steeper along the sequence of stellar evolution phases,

65: with exponents $0.7\pm 0.2$, $1.3\pm 0.15$, and $2.2\pm 0.2$ for Class~0, 

66: Class~I, and Class~II systems, respectively.

67: \end{abstract}

68: 

69: \keywords{circumstellar matter --- planetary systems: protoplanetary disks --- hydrodynamics --- ISM: clouds ---  stars: formation} 

70: 

71: 

72: 

73: \section{Introduction}

74: It has now become evident that disks of gas and dust are present from the earliest 

75: phases of stellar evolution (Class~0 and Class~I) and last for at least several million years

76: into the late Class~II phase. Disks are observed or inferred around most T Tauri stars and 

77: even around brown dwarfs. 

78: The evidence for disks around Class~0 and Class~I sources is more indirect.

79: In this early phase of stellar evolution, the protostar/disk system is deeply 

80: embedded in an envelope -- a remnant of the cloud core from which the protostar is forming. 

81: Nevertheless, recent observations by \cite{Andrews} suggest that Class I disks

82: have a larger median mass than that of Class II disks.

83: 

84: 

85: In spite of a considerable progress in the detection of disks around young stellar objects 

86: (YSOs), an accurate determination of disk masses is still challenging.

87: It is difficult to directly determine disk masses from the spectral lines 

88: of molecular species because the brightest, easily detectable lines 

89: (i.e., the rotational transitions of CO) are optically thick and likely to be severely 

90: depleted. Therefore, disk masses are usually inferred from analyzing the spectral 

91: energy distribution of YSOs from the mid-infrared through submillimeter bands 

92: \citep[e.g.][]{Andrews,Brown}. Such measurements of disk masses suffer from large 

93: uncertainties in the normalization of dust opacities and gas-to-dust ratios,

94: which led \citet{Hartmann} to conclude that T Tauri disk masses have been 

95: systematically underestimated by conventional analyses.

96: 

97: Another complication arises from poorly known physical processes in the disk.

98: A usual assumption of optically thin circumstellar disks may significantly 

99: underestimate disk masses, particularly for objects with larger flux densities.

100: However, a self-consistent treatment of a non-negligible optical depth requires 

101: a knowledge of the radial gas surface density profile in the disk \citep{Andrews}, 

102: which may depend significantly on the dominant physical mechanism of mass and angular 

103: momentum redistribution in the disk \citep{VB4}.

104: 

105: Given large uncertainties in the measurements of disk masses, numerical simulations

106: of self-consistent formation and evolution of circumstellar disks can provide  

107: valuable information on disk masses in the early embedded and late phases 

108: of YSO evolution. It has been shown in the past that the saturation of spiral

109: gravitational instabilities at a finite amplitude in a self-gravitating, Toomre-unstable disk 

110: allows for the steady transport of mass and momentum, which eventually limits

111: disk masses \citep[e.g.][]{Adams89,Shu90,Laughlin97,Laughlin98}.

112: In this paper, we perform numerical simulations of

113: the long-term evolution of self-consistently formed circumstellar disks

114: around low-mass stars ($0.2~M_\odot \la M_\ast\la 2.0~M_\odot$).

115: We consider both the self-gravitating disks, in which radial transport of mass and

116: angular momentum is done exclusively via gravitational torques, and viscous disks, which feature 

117: gravitational torques as well as viscous ones. 

118: We seek to determine numerically the disk masses in the Class 0, Class I, and Class II phases

119: of stellar evolution.

120: 

121: The paper is organized as follows. 

122: The numerical methods and initial parameters of cloud cores are 

123: given in \S~\ref{methods}. Our obtained masses for self-gravitating and viscous disks are

124: presented in \S~\ref{SG} and \S~\ref{VD}. We compare our numerical results with observations in 

125: \S~\ref{discuss}. The model and numerical caveats are discussed in \S~\ref{caveats}. 

126: The main results are summarized in \S~\ref{summary}.

127: 

128: 

129: \section{Description of numerical methods}

130: \label{methods}

131: We use the thin-disk approximation to compute the evolution of rotating, 

132: gravitationally bound cloud cores. This allows efficient calculation of 

133: the long-term evolution of a large number of models.

134: We start our numerical integration in the pre-stellar phase, which is characterized 

135: by a collapsing {\it starless} cloud core, and continue into the main accretion phase,

136: which sees the formation of a central star and circumstellar disk.

137: We cover all major phases of the evolution of a YSO, starting from 

138: its formation and ending with the T Tauri phase. The integration ends when the age of 

139: the central star is about three million years. In some models we extend the integration

140: up to 5~Myr. We emphasize that circumstellar disks

141: are formed self-consistently in our numerical simulations, 

142: rather than being introduced as an initial parameter of the model.

143: 

144: 

145: Once the disk is formed, its mass is determined by an interplay between the efficiency 

146: of the mass and angular momentum transport in the disk\footnote{In fact, disks may also transport

147: angular momentum to the external environment due to magnetic braking. This effect will be 

148: considered in a follow-up paper.} and the infall rate of matter from the surrounding envelope onto the

149: disk. At the time of disk formation, the infall rates take values between 

150: $1.2\times 10^{-6}~M_\odot$~yr$^{-1}$ and  $7.0\times 10^{-6}~M_\odot$~yr$^{-1}$  (measured at $600$~AU)

151: for the least and most massive cloud cores, respectively, but they show a fast decline with time.

152: These values (and strong time variation) are consistent with the infall rates 

153: derived by \citet{Klessen01} using 

154: numerical models that follow molecular cloud evolution from turbulent fragmentation toward the 

155: formation of stellar clusters. We note that once the disk is formed, 

156: the infall rate of matter from the envelope onto the disk is not necessarily the same 

157: as the mass accretion rate from the disk onto the protostar. While the former shows a fast decline

158: with time, the latter is usually characterized by a much slower decline and has a strong dependence

159: on the stellar mass \citep{VB3,VB5}.

160: 

161: 

162: We use two numerical approaches: a basic approach that accounts for the radial transport due to gravitational

163: toques and a viscous approach that accounts for the radial transport due to both the gravitational

164: torques and viscosity.

165: Gravitational torques are known to efficiently redistribute mass and 

166: angular momentum in circumstellar disks \citep[e.g.][]{Lodato04,Lodato05}. 

167: They were shown to play an important role in driving the FU-Ori-like bursts

168: in the early embedded phase of disk evolution \citep{VB1,VB2}. In the late disk evolution, 

169: negative gravitational torques associated with low-amplitude azimuthal density perturbations

170: in the disk can drive mass accretion rates that are consistent with those measured in 

171: the intermediate and upper-mass T Tauri stars \citep{VB3,VB5}.

172: 

173: 

174: 

175: \subsection{Basic numerical approach}

176: \label{basic}

177: 

178: In the basic numerical approach, the collapse of a cloud core and subsequent evolution 

179: of a star/disk system 

180: is carried out by solving the basic equations of mass and momentum transport 

181: in the thin-disk approximation \citep[see e.g.][]{VB2}

182: %are derived by integrating the ideal hydrodynamics equations in the z-direction,

183: %from $z = Z(r, \phi, t)$ to $z = -Z(r, \phi, t)$, and using Leibniz's rule

184: %for differentiation of integrals and the fundamental theorem of

185: %calculus \citep[see e.g.][]{VB2}

186: \begin{eqnarray}

187: \label{cont}

188:  \frac{{\partial \Sigma }}{{\partial t}} & = & - \nabla _p  \cdot \left( \Sigma \bl{v}_p 

189: \right), \\ 

190: \label{mom}

191:  \Sigma \frac{d \bl{v}_p }{d t}  & = &  - \nabla _p {\cal P}  + \Sigma \, \bl{g}_p \, ,

192: \end{eqnarray}

193: where $\Sigma$ is the mass surface density, ${\cal P}=\int^{Z}_{-Z} P dz$ is the vertically integrated

194: form of the gas pressure $P$, $Z$ is the radially and azimuthally varying vertical scale height,

195:  $\bl{v}_p=v_r \hat{\bl r}+ v_\phi \hat{\phi}$ is the velocity in the

196: disk plane, $\bl{g}_p=g_r \hat{\bl r} +g_\phi \hat{\phi}$ is the gravitational acceleration 

197: in the disk plane, and $\nabla_p=\hat{\bl r} \partial / \partial r + \hat{\phi} r^{-1} 

198: \partial / \partial \phi $ is the gradient along the planar coordinates of the disk. 

199: The gravitational acceleration $\bl{g}_p$ is found by solving for the Poisson integral 

200: \citep[see][]{VB2}. The fact that we account for the disk self-gravity means

201: that gravitational torques arise {\it self-consistently} in our numerical simulations and

202: not imitated by some means of $\alpha$-viscosity. 

203: 

204: Taking into account the complexity of gas thermodynamics in circumstellar disks

205: (see \S~\ref{caveats}), we have adopted a barotropic 

206: equation of state that closes equations~(\ref{cont}) and (\ref{mom}) and makes 

207: a transition from isothermal to adiabatic evolution at $\Sigma = \Sigma_{\rm cr} = 

208: 36.2$~g~cm$^{-2}$

209: \begin{equation}

210: {\cal P}=c_s^2 \Sigma +c_s^2 \Sigma_{\rm cr} \left( \Sigma \over \Sigma_{\rm cr} \right)^{\gamma},

211: \label{barotropic}

212: \end{equation}

213: where $c_s$ is the isothermal sound speed, the value of which is set equal to that 

214: of the initial cloud core, and $\gamma=1.4$. 

215: Equation~(\ref{barotropic}), though neglecting detailed

216: cooling and heating processes, was shown to reproduce to a first approximation 

217: the radial temperature gradients in the disk \citep{VB3} and the density-temperature relation 

218: for collapsing cloud cores derived by \citet{Masunaga} using a detailed radiation hydrodynamics 

219: simulation \citep{VB2}.

220: 

221: The vertical scale height $Z(r,\phi,t)$ is determined assuming the local hydrostatic 

222: equilibrium in the gravitational field of a disk and central star \citep{VB4}.

223: The relevant formulas are given in the Appendix.

224: 

225: \subsection{Viscous numerical approach}

226: \label{viscmodel}

227: Viscosity is another important mechanism of angular momentum 

228: and mass redistribution in astrophysical disks. Most analytical and numerical studies 

229: of viscous evolution of thin circumstellar disks have employed 

230: the standard axisymmteric model of \citet{Lyndenbell}, in which 

231: the surface density of a Keplerian disk evolves with time according to the following diffusion equation

232: \begin{equation}

233: {\partial \Sigma \over \partial t} = {3\over r} {\partial \over \partial r} 

234: \left[ r^{1/2} {\partial \over \partial r} ( \nu \Sigma r^{1/2}) \right],

235: \end{equation}

236: where $\nu$ is the kinematic viscosity. 

237: 

238: In the present paper, we take a more fundamental approach and describe the effect of (yet unspecified)

239: viscosity in terms of the classic viscous stress tensor 

240: \begin{equation}

241: \mathbf{\Pi}=2 \mu \left( \nabla v - {1 \over 3} (\nabla \cdot v) \mathbf{e} \right),

242: \end{equation}

243: where $\nabla v$ is a symmetrized velocity gradient tensor, $\mathbf{e}$ is the unit tensor, and

244: $\mu$ is the dynamical viscosity.

245: This approach allows for a 

246: self-consistent treatment of both, self-gravity and viscosity, within the same numerical formalism.

247: The resulting mass and momentum transport equations in the viscous numerical approach are

248: \begin{eqnarray}

249: \label{visc1}

250:  \frac{{\partial \Sigma }}{{\partial t}} & = & - \nabla _p  \cdot \left( \Sigma \bl{v}_p 

251: \right), \\ 

252: \label{visc2}

253:  \Sigma \frac{d \bl{v}_p }{d t}  & = &  - \nabla _p {\cal P}  + \Sigma \, \bl{g}_p

254:  + \left( \nabla \cdot \mathbf{\Pi}\right)_p \, ,

255: \end{eqnarray}

256: where $\nabla \cdot \mathbf{\Pi}$ is the divergence of the rank-two viscous stress tensor 

257: $\mathbf{\Pi}$. The relevant components of $\left( \nabla \cdot \mathbf{\Pi} \right)_p$ 

258: are given in the Appendix.

259: Equations~(\ref{visc1}) and (\ref{visc2}) are closed with the barotropic equation of state~(\ref{barotropic}).

260: We note that the viscous approach accounts self-consistently for both the gravitational and 

261: viscous torques which may arise during numerical simulations.

262: 

263: It is evident that the practical application of equation~(\ref{visc2}) requires a knowledge

264: of the dynamical viscosity $\mu$ of the disk.

265: Unfortunately, our understanding of viscous processes in circumstellar disks is still 

266: incomplete. We know that molecular (collisional) viscosity is most certainly too low to be of 

267: practical interest. Turbulence driven by the magneto-rotational instability (MRI) 

268: is a most promising source of viscosity at present \citep{BH}, though other

269: mechanisms cannot be ruled out completely.

270: 

271: In this paper, we make no specific assumptions as to the source of viscosity and

272: define the coefficient of dynamical viscosity using the usual $\alpha$-prescription

273: of \citet{SS}

274: \begin{equation}

275: \mu=\alpha \, \Sigma \, \tilde{c}_s \, Z,

276: \end{equation} 

277: where spatially and temporally uniform $\alpha$ is set equal to 0.01 and 

278: $\tilde{c}_s=\sqrt{\partial {\cal P}/ \partial \Sigma}$ is the effective

279: sound speed. Our numerical simulations of embedded and T Tauri disks

280: indicate that $\alpha\simeq 0.001-0.01$ yields disk sizes and radial slopes 

281: that are in general agrement with observations \citep{VB4}. Lower values of $\alpha$ 

282: ($ < 10^{-3}$) have little effect on the evolution of self-gravitating disks, 

283: whereas substantially higher values ($\ga 0.1$) quickly destroy the disks.

284: \citet{Hartmann98} also predicted similar values for $\alpha$ by

285: analyzing accretion rates in T Tauri disks.

286: Since viscosity in our  model is assumed to arise due to some physical processes 

287: {\it in the disk}, we keep $\alpha$ equal to zero during the early ``pre-disk'' phase of 

288: evolution and set $\alpha$ equal to 0.01 only when a circumstellar disk forms

289: around a central star. 

290: 

291: 

292: 

293: 

294: \subsection{Initial condition}

295: 

296: The initial radial distributions of surface density $\Sigma$ and angular velocity $\Omega$ 

297: in our model cloud cores

298: are those characteristic of a collapsing axisymmetric magnetically

299: supercritical core \citep{Basu}

300: \begin{equation}

301: \Sigma={r_0 \Sigma_0 \over \sqrt{r^2+r_0^2}}\:,

302: \label{dens}

303: \end{equation}

304: \begin{equation}

305: \Omega=2\Omega_0 \left( {r_0\over r}\right)^2 \left[\sqrt{1+\left({r\over r_0}\right)^2

306: } -1\right],

307: \end{equation}

308: where $r_0$ is the radial scale length defined as $r_0 = k c_s^2 /(G\Sigma_0)$ and  $k= \sqrt{2}/\pi$.

309: These initial profiles are characterized by the important

310: dimensionless free parameter $\eta \equiv  \Omega_0^2r_0^2/c_s^2$

311: and have the property 

312: that the asymptotic ($r \gg r_0$) ratio of centrifugal to gravitational

313: acceleration has magnitude $\sqrt{2}\,\eta$ \citep[see][]{Basu}. 

314: The centrifugal radius of a mass shell initially located at radius $r$ is estimated to be

315: $r_{\rm cf} = j^2/(Gm) = \sqrt{2}\, \eta r$, where $j=\Omega r^2$ is the specific angular

316:  momentum. Since the enclosed mass $m$ is a linear function of $r$ at large radii,

317: this also means that $r_{\rm cf} \propto m$.

318: The gas has a mean molecular mass $2.33 \, m_{\rm H}$ and cloud cores are initially

319: isothermal with temperature $T=10$ K. 

320: 

321: 

322: 

323: We present results from three sets of models, each with a different value of $\eta$.  

324: The standard model has $\eta = \eta_1 = 1.2 \times 10^{-3}$ based on typical values

325: $c_s = 0.19$ km s$^{-1}$, $\Sigma_0 = 0.12$~g~cm$^{-2}$, and

326: $\Omega_0 = 1.0$~km~s$^{-1}$~pc$^{-1}$. The outer radius is taken to

327: be $r_{\rm out} = 0.04$ pc, and the total cloud mass is $0.8\,\msun$. 

328: Other models with $\eta = \eta_1$ but different mass (outer radius) 

329: are generated by varying $r_0$ and $\Omega_0$ so that their product is constant. 

330: Note that, when $r_0$ is varied, $\Sigma_0$ has to be changed accordingly.

331: All clouds are characterized by the same ratio $r_{\rm out}/r_0\approx 6.0$. 

332: To generate the second set of models, $\eta = \eta_2 = 2.3 \times 10^{-3}$, we set

333: $\Omega_0 = 1.4$~km~s$^{-1}$~pc$^{-1}$ and all other quantities the

334: same as in the standard model with $\eta= \eta_1$. Models of

335: varying mass are then generated in the same manner as for the

336: $\eta_1$ models. The third set of models, with $\eta=\eta_3

337: = 3.4 \times 10^{-3}$, are also obtained in this way,

338: by first using $\Omega_0 = 1.7$~km~s$^{-1}$~pc$^{-1}$.

339: Overall, there are 7 models with $\eta = \eta_1$, 13 models with

340: $\eta = \eta_2 \simeq 2\, \eta_1$, 

341: and 12 with $\eta = \eta_3 \simeq 3\, \eta_1$.

342: The range of initial cloud masses ($M_{\rm cl}$) amongst our models is $0.3\,\msun-2.95\,\msun$.

343: The parameters of our models are listed in Table~\ref{table1} ($\eta_1$), Table~\ref{table2}

344: ($\eta_2$), and Table~\ref{table3} ($\eta_3$).

345: We note that our model values of $\Omega_0 = (1.0-1.7)$~km~s$^{-1}$~pc$^{-1}$ are within a typical 

346: range of velocity gradients measured in dense starless cores by \citet{Caselli}.

347: 

348: 

349: \begin{table}

350: \begin{center}

351: \caption{Parameters of models with $\eta_1=1.2 \times 10^{-3}$

352: \label{table1}}

353: %\vspace{3 pt}

354: \begin{tabular}{clllll}

355: \hline\hline

356: Model & $r_0$ & $\Sigma_0$ & $\Omega_0$ & $r_{\rm out}$ & $M_{\rm cl}$  \\

357: \hline

358:  1 & 1209 & 0.13  & 1.14 & 7186  & 0.7 \\

359:  2 & 1382 & 0.12  & 1.0  & 8213  & 0.8 \\

360:  3 & 1728 & 0.093 & 0.8  & 10266 & 0.98 \\

361:  4 & 2074 & 0.077 & 0.67 & 12320 & 1.18 \\

362:  5 & 2937 & 0.055 & 0.47 & 17452 & 1.67 \\

363:  6 & 4147 & 0.039 & 0.33 & 24640 & 2.36 \\

364:  7 & 5184 & 0.031 & 0.27 & 30800 & 2.95 \\

365:  \hline

366: \end{tabular} 

367: \tablecomments{All distances are in AU, angular

368: velocities in km~s$^{-1}$~pc$^{-1}$, surface densities in g~cm$^{-2}$, and masses in $M_\odot$.}

369: \end{center}

370: \end{table} 

371: 

372: \begin{table}

373: \begin{center}

374: \caption{Parameters of models with $\eta_2=2.3 \times 10^{-3}$

375: \label{table2}}

376: %\vspace{3 pt}

377: \begin{tabular}{clllll}

378: \hline\hline

379: Model & $r_0$ &  $\Sigma_0$ & $\Omega_0$ & $r_{\rm out}$ & $M_{\rm cl}$  \\

380: \hline

381:  8 & 622  & 0.26  & 3.1  & 3696  & 0.35 \\

382:  9 & 691  & 0.23  & 2.8  & 4106  & 0.4 \\

383:  10 & 864  & 0.19  & 2.24 & 5133  & 0.5 \\

384:  11 & 1037 & 0.16  & 1.87 & 6160  & 0.6 \\

385:  12 & 1210 & 0.13  & 1.6  & 7187  & 0.7 \\

386:  13 & 1417 & 0.11  & 1.4  & 8213  & 0.8  \\

387:  14 & 1728 & 0.093 & 1.12 & 10267 & 0.98 \\

388:  15 & 2454 & 0.065 & 0.79 & 14579 & 1.4  \\

389:  16 & 2765 & 0.058 & 0.7  & 16428 & 1.57 \\

390:  17& 3456 & 0.046 & 0.56 & 20533 & 1.97 \\

391:  18& 3802 & 0.042 & 0.51 & 22587 & 2.16 \\

392:  19& 4147 & 0.038 & 0.47 & 24640 & 2.36 \\

393:  20& 4838 & 0.033 & 0.4  & 28747 & 2.75 \\

394: % 14& 5530 & 0.029 & 0.35 & 32853 & 3.15 \\

395:  \hline

396: \end{tabular} 

397: \tablecomments{All distances are in AU, angular

398: velocities in km~s$^{-1}$~pc$^{-1}$, surface densities in g~cm$^{-2}$, and masses in $M_\odot$.}

399: \end{center}

400: \end{table} 

401: 

402: 

403: \begin{table}

404: \begin{center}

405: \caption{Parameters of models with $\eta_2=3.4 \times 10^{-3}$

406: \label{table3}}

407: %\vspace{3 pt}

408: \begin{tabular}{clllll}

409: \hline\hline

410: Model & $r_0$ & $\Sigma_0$ & $\Omega_0$ & $r_{\rm out}$ & $M_{\rm cl}$  \\

411: \hline

412:  21 & 518 & 0.31  & 4.53  & 3080  & 0.3 \\

413:  22 & 691  & 0.23  & 3.4  & 4106  & 0.4 \\

414:  23 & 864  & 0.19  & 2.7  & 5133  & 0.5 \\

415:  24 & 1037 & 0.16  & 2.26 & 6160  & 0.6 \\

416:  25 & 1210 & 0.13  & 1.94 & 7187  & 0.7 \\

417:  26 & 1417 & 0.11  & 1.7  & 8213  & 0.8  \\

418:  27 & 2073 & 0.077 & 1.13 & 12320 & 1.18 \\

419:  28 & 2420 & 0.066 & 0.97 & 14373 & 1.38 \\

420:  29 & 3283 & 0.049 & 0.72 & 19506 & 1.87  \\

421:  30 & 3802 & 0.042 & 0.62  & 22587 & 2.16 \\

422:  31 & 4147 & 0.039 & 0.57 & 24640 & 2.36 \\

423:  32 & 4493 & 0.036 & 0.52 & 26693 & 2.56 \\

424: % 32& 5530 & 0.029 & 0.43 & 32853 & 3.15 \\

425: % 13& 4838 & 0.033 & 0.4  & 28747 & 2.75 \\

426: % 14& 5530 & 0.029 & 0.35 & 32853 & 3.15 \\

427:  \hline

428: \end{tabular} 

429: \tablecomments{All distances are in AU, angular

430: velocities in km~s$^{-1}$~pc$^{-1}$, surface densities in g~cm$^{-2}$, and masses in $M_\odot$.}

431: \end{center}

432: \end{table} 

433: 

434: 

435: \subsection{Numerical technique}

436:  

437: Hydrodynamic equations of the basic and viscous numerical models are solved in polar 

438: coordinates $(r, \phi)$ on a numerical grid with

439: $128 \times 128$ points. We use the method of finite differences with a time-explicit,

440: operator-split solution procedure. Advection is

441: performed using the second-order van Leer scheme.  The radial points are logarithmically spaced.

442: The innermost grid point is located at $r=5$~AU, and the size of the 

443: first adjacent cell lies in the range between 0.26~AU (model~21) 

444: and 0.36~AU (model~7). It means that the ratio $\Delta r/r$ of the cell size $\Delta r$ 

445: to radius $r$ is constant for a given cloud core and varies from $0.05$ (model~21) to 

446: $0.07$ (model~7). 

447: %It can be noticed from Fig.~\ref{figA1} that the aspect ratio $A=z/r$

448: %of the disk vertical scale height to radius is smaller than $\Delta r/r$ in the inner 70~AU, 

449: %which implies that a better radial (and angular) resolution is desirable in this inner region. 

450: %We have found that an increase in the resolution to $256 \times 256$ grid points makes 

451: %little influence on the accretion history and disk masses but helps save a considerable amount 

452: %of CPU time and, eventually, consider many more model cloud cores. We also point out that, due to

453: %the adopted log spacing in the radial direction, the resolution in the inner computational

454: %region is sufficient to fulfill the Truelove criterion in the disk.

455: 

456:  We introduce a ``sink cell'' at $r<5$~AU, 

457: which represents the central star plus some circumstellar disk material, 

458: and impose a free inflow inner boundary condition. About 95 per cent of 

459: the material crossing the inner boundary lands into the star, the

460: rest constitutes an inner circumstellar region of constant surface density.     

461: The dynamics of the inner region ($r<5$~AU) is not computed but it contributes 

462: to the total gravitational potential of the system. 

463: We do not account for a possible mass loss from the sink cell due to stellar jets, 

464: since our model is two-dimensional and

465: it is not clear how the mass ejection efficiency varies with stellar age. 

466: The outer boundary is such that the cloud has a constant mass and volume. 

467: More details on numerical techniques and relevant tests are given in \citet{VB2}.

468: 

469: 

470: 

471: 

472: \section{Masses of self-gravitating disks}

473: \label{SG}

474: In this section we present disk masses obtained using the basic numerical approach

475: described in \S~\ref{basic}. In the framework of this numerical model, disk masses 

476: are controlled by the rate of mass infall from 

477: the surrounding envelope and the radial transport of mass and angular momentum 

478: due to gravitational torques. No viscous torques are present in this case.

479: 

480: An accurate determination of disk masses in numerical simulations of collapsing cloud cores is not

481: a trivial task. Self-consistently formed circumstellar disks have a wide range of masses 

482: and sizes, which are not known {\it a priori}. However, numerical and observational 

483: studies of circumstellar disks indicate that the disk

484: surface density is a declining function of radius. Therefore, we distinguish

485: between disks and infalling envelopes using a critical surface density for the 

486: disk-to-envelope transition, for which we choose a value of $\Sigma_{\rm tr}=0.1$~g~cm$^{-2}$.

487: This choice is dictated by the fact that densest {\it starless} cores have surface

488: densities only slightly lower than the adopted value of $\Sigma_{\rm tr}$.

489: In addition, our numerical simulations indicate that self-gravitating disks have 

490: sharp outer edges and the gas densities of order $0.01-0.1$~g~cm$^{-2}$ 

491: characterize a typical transition region between the disk and envelope \citep{VB3}.

492: 

493: 

494: 

495: To compare disk masses in three distinct phases of stellar evolution

496: we need an evolutionary indicator to distinguish between Class 0, Class I, 

497: and Class II phases. 

498: We use a classification of \citet{Andre}, who

499: suggest that the transition between Class 0 and Class I objects occurs when

500: about $50\%$ of the initial cloud core is accreted onto the protostar/disk system.

501: The Class II phase is consequently defined by the time when the

502: infalling envelope clears and its total mass drops below 10\% of the initial cloud 

503: core mass $M_{\rm cl}$.

504: It should be mentioned here that there exists no unique classification scheme for protostars.

505: For instance, \citet{VB0} have proposed a classification scheme that hinges 

506: on a temporal behaviour 

507: of bolometric luminosity $L_{\rm bol}$. They identify Class~0 phase with a period of 

508: temporally increasing $L_{\rm bol}$ and Class~I phase with a later period of decreasing 

509: $L_{\rm bol}$. The peak in $L_{\rm bol}$ corresponds to the evolutionary time when 

510: $50 \pm 15$ per cent of the cloud core mass has been accreted by the protostar.

511: Observers prefer to classify protostars by their spectral energy distributions. 

512: For instance, \citet{Lada87} and \citet{Andrews} use the values of power-law index $n$ 

513: (defined by $\nu F_{\nu} \propto \nu^n$, where $F_{\nu}$ is the infrared flux density at frequency 

514: $\nu$) to distinguish between Class~I and Class~II objects. 

515: \citet{Andre} use the ratio of submillimeter to bolometric luminosity

516: $L_{\rm submm}/L_{\rm bol}$ for the same purposes. Although these classifications are

517: physically related, it is possible that they may differ from our adopted classification, 

518: especially if some of the gas is removed

519: from the cloud core and does not accrete. We acknowledge that the difference 

520: in the existing classification schemes may systematically shift our results. 

521: 

522: 

523: \subsection{Temporal evolution of self-gravitating disks}

524: \label{tempevol}

525: We start by comparing the long-term evolution of disk masses in four sample models, which are 

526: chosen to represent a full spectrum of the initial cloud core 

527: masses with a constant ratio $\eta=2.3\times 10^{-3}$ of centrifugal to gravitational 

528: acceleration. Figure~\ref{fig1} shows the disk masses (thick solid lines), stellar masses (thin

529: solid lines), and envelope masses (dashed lines) in model~8 (upper left, $M_{\rm cl}=0.35~M_\odot$),

530: model~11 (upper right, $M_{\rm cl}=0.6~M_\odot$), model~16 (lower left, $M_{\rm cl}=1.57~M_\odot$),

531: and model~20 (lower right, $M_{\rm cl}=2.75~M_\odot$). The horizontal axis is the time elapsed 

532: since the formation of a central star.

533: All masses are calculated relative to the corresponding 

534: initial cloud core mass $M_{\rm cl}$. The vertical dotted lines mark the onset 

535: of Class I (left)  and Class~II (right) phases.

536: 

537: \begin{figure}

538:   \resizebox{\hsize}{!}{\includegraphics{f1.eps}}

539: % \centering

540: %  \includegraphics{figure1.eps}

541:       \caption{Disk masses (thick solid lines), stellar masses (thin solid lines),

542:       and envelope masses (dashed lines) obtained in the framework of 

543:       the basic numerical approach (see \S~\ref{basic}) 

544:       for model~8 (upper left), model~11 (upper right), 

545:       model~16 (lower left), and model~20 (lower right).

546:       All masses are calculated relative to the corresponding 

547:       initial cloud core mass $M_{\rm cl}$. The horizontal axis shows time elapsed since the 

548:       formation of a central star. The left/right vertical dotted lines mark the onset 

549:       of Class I/II phases, respectively.}

550:          \label{fig1}

551: \end{figure}

552: 

553: Our numerical simulations demonstrate that cloud cores with constant $\eta$ (but

554: different size) form disks roughly at the same physical time after the formation 

555: of a central star but in distinct stellar evolution phases. For instance,

556: model~8 starts to build a disk in the late Class I phase, while model~20 does that in the midst

557: of Class 0 phase. Cloud cores of greater mass tend to form disks in the earlier phase of stellar 

558: evolution than their low-mass counterparts. 

559: 

560: The upper panel of Fig.~\ref{fig1} identifies cases when Class~0 stars have no disks

561: associated with them.  

562: We emphasize here that due to the use of the sink cell in our numerical code

563: we can resolve only those disks whose outer radii are larger than 5~AU. It means that even though 

564: circumstellar disks do not form around some Class 0 objects in our numerical simulations, 

565: as in models~8 and 11 

566: (top panels in Fig.~\ref{fig3}), one may suppose that such disks still exist  

567: but their size is simply smaller than that of the sink cell. 

568: Our test runs with a sink cell set to 0.5~AU (instead of 5~AU) confirm

569: that disks indeed forms earlier in the evolution but quickly expands to 5~AU and beyond.

570: Hence, the absence of disks around some Class~0 objects may be to some extent 

571: caused by a finite-size sink cell. However, it is still possible that some Class~0 

572: objects have no disks associated with them. This is particularly true for those objects that 

573: develop from cloud cores with low rates of rotation, since in this case

574: a (substantial) portion of the disk material will be characterized

575: by the centrifugal radius (estimated as $r_{\rm cf}=j^2/(Gm)$) that is

576: smaller than the radius of the protostar, $4~R_\odot$.

577: A numerical code with realistic treatment of the protostar formation is needed

578: to accurately address this issue.

579: 

580: 

581: Another interesting feature of self-gravitating disks that can be seen in Fig.~\ref{fig1} 

582: is that the disk-to-star mass 

583: ratio never exceeds some characteristic value, approximately $0.35-0.4$, irrespective of the

584: initial cloud core mass. This is rather counterintuitive. For instance, models~16 and 20 (bottom panels

585: in Fig.~\ref{fig1}) form disks in the Class 0 phase, which is characterized by envelope 

586: masses that are considerably greater than those of the protostars. As a consequence,

587: one may expect the formation of a disk with mass that is at least comparable 

588: to or even greater than that of a protostar. It turns out, however, that circumstellar disks 

589: that form in the {\it Class~0 or early Class~I phase} around stars with mass 

590: $M_\ast \ga 0.6~M_\odot$ develop vigourous gravitational 

591: instability -- a very efficient means of inward mass transport 

592: that helps keep the disk mass well below that of the protostar. 

593: Sharp drops in the disk mass (or equivalent surges in the stellar mass) 

594: seen in the upper right and bottom panels of Figure~\ref{fig1} are a manifestation of 

595: this process \citep[see e.g.][]{VB1,VB2}.  

596: \citet{Kratter} also predict that disks around stars with mass greater than $1.0~M_\odot$

597: are expected to be vigorously gravitationally unstable in the early embedded phase of

598: disk evolution. 

599: The late evolution phase (Class II) sees only an insignificant decline of disk mass with time. 

600: 

601: 

602: \begin{figure}

603:   \resizebox{\hsize}{!}{\includegraphics{f2.eps}}

604:       \caption{Time-averaged disk masses (right column) and

605:       time-averaged disk-to-star mass ratios (left column) versus time-averaged stellar masses. 

606:       Open triangles, filled squares, and plus signs show the data for Class~0, Class~I, 

607:       and Class~II 

608:       systems, respectively. In particular, top panels show the data for models 

609:       with $\eta_1 = 1.2 \times 10^{-3}$, whereas middle and bottom panels show the data for 

610:       models with $\eta_2\simeq 2\eta_1$ and $\eta_3\simeq 3\eta_1$, respectively.}

611:          \label{fig2}

612: \end{figure}

613: 

614: 

615: \subsection{Disk masses versus stellar masses}

616: We use 32 models, the parameters of which are listed in 

617: Tables~\ref{table1}-\ref{table3}, to analyse the statistical relations between time-averaged 

618: disk and stellar masses in the Class 0/I/II evolution phases. 

619: Time averaging is performed separately for each evolution phase over the duration of the phase.

620: The age of the oldest disk in our sample 

621: is about 3~Myr. Since disk masses in the Class II phase have a tendency to decline with 

622: time and the actual

623: lifetime of Class II objects may be longer, we expect that

624: we might have somewhat overestimated disk masses around Class II objects. This, obviously, does not

625: effect our estimates of Class 0/I disk masses. 

626: 

627: Figure~\ref{fig2} shows time-averaged disk masses (right column) and time-averaged 

628: disk-to-star mass ratios (left column) versus time-averaged stellar masses 

629: for Class~0 (open triangles), Class~I (filled squares) and Class~II  objects (plus signs).

630:  Several interesting conclusions can be drawn 

631: by analysing the figure. 

632: \begin{enumerate}

633:  \item Time-averaged disk and stellar masses ($\langle M_{\rm d}\rangle$ and 

634:  $\langle M_\ast \rangle$, 

635:  respectively) in models with the same ratio of rotational to gravitational 

636: acceleration fall onto a unique evolutionary track in the $\langle M_{\rm d} \rangle - \langle M_\ast

637: \rangle$  diagram. 

638: \item  Class 0 objects occupy the lower-left part of each evolutionary track. No stars with 

639: $\langle M_\ast \rangle \ga 0.9~M_\odot$ have Class 0 disks (but they have Class~I/II disks). 

640: In other words, only low-mass stars (within our range of interest, 

641: $0.2~M_\odot \la \langle M_\ast \rangle  \la 2.0~M_\odot$) can harbour Class~0 disks.

642: \item Stars of {\it equal mass} have disks with similar masses, regardless of the stellar 

643: evolution phases. 

644: \item In each stellar evolution phase, disk-to-star mass ratios tend to have greater values 

645: for stars of greater mass. However, there is a clear saturation effect for stars with

646: masses greater than $1.0~M_\odot$ -- disk masses never grow above $40\%$ 

647: of the stellar masses, even for models with the largest values of $\eta = \eta_3=3.4\times 10^{-3}$.

648: The saturation of disk-to-star mass ratios is caused by the onset of vigorous

649: gravitational instability in circumstellar disks that form in the Class~0 and early Class~I phase.

650: \end{enumerate}

651: 

652: 

653: \begin{table*}

654: \begin{center}

655: \caption{Summary of disk properties}

656: \label{table4}

657: %\vspace{3 pt}

658: \begin{tabular}{ccccccc}

659: \hline

660: \hline

661: disk type & $ \!\!\! \overline{M}_{\rm d,C0}$ & $\!\!\! \overline{M}_{\rm d,CI}$ & $\!\!\! \overline{M}_{\rm d,CII}$ & 

662: $\!\!\! \left( \langle M^{\rm min}_{\rm d,C0}\rangle : \langle M^{\rm max}_{\rm d,C0} \rangle \right)$ & 

663: $ \!\!\! \left( \langle M^{\rm min}_{\rm d,CI} \rangle : \langle M^{\rm max}_{\rm d,CI} \rangle \right) $

664:  &  $ \!\!\! \left( \langle M^{\rm min}_{\rm d,CII} \rangle : \langle M^{\rm max}_{\rm d,CII} 

665:  \rangle \right)$ \\

666: \hline

667: self-gravitating& 0.09 & 0.10 & 0.12  & $(0.04 : 0.2)$ & $(0.02 : 0.5)$  & $(0.03 : 0.7)$ \\

668: viscous & 0.10 & 0.11 & 0.06  & $(0.05 : 0.19)$  & $(0.02 : 0.55)$  & $(0.004 : 0.6)$ \\

669:  \hline

670: \end{tabular} 

671: \tablecomments{Mean masses $\overline{M}_{\rm d,C0}$, $\overline{M}_{\rm d,CI}$, and 

672: $\overline{M}_{\rm d,CII}$ of Class 0, Class I, and Class II disks, respectively, are

673: calculated according to equation~(\ref{mean}). Angle brackets $\langle \, \rangle$ denote

674: time averaging over the duration of a particular stellar evolution phase. 

675: Indices ``min'' and ``max'' refer to minimum and maximum time-averaged disk masses, respectively. 

676: All disk masses are in $M_\odot$.}

677: \end{center}

678: \end{table*} 

679: 

680: 

681: 

682: We summarize the main properties of our model self-gravitating disks in Table~\ref{table4}.

683: In order to facilitate the comparison of our numerical results with 

684: observations, we calculate mean disk masses (in each evolution phase) 

685: by weighting the time-averaged disk masses according to the corresponding stellar masses.

686: The relative importance of the stellar masses is calculated using the following power-law 

687: weight function

688: \begin{equation}

689: {\cal F}_{\rm K}\left(\langle M_\ast \rangle\right) = \left\{ \begin{array}{ll} 

690:    A \, \langle M_\ast \rangle^{-1.3} & \,\, 

691:    \mbox{if $\langle M_\ast \rangle \le 0.5~M_\odot$ } \\ 

692:    B \, \langle M_\ast \rangle^{-2.3} & \,\, 

693:    \mbox{if $\langle M_\ast \rangle > 0.5~M_\odot$ }.  \end{array} 

694:    \right. 

695:    \label{function}  

696:  \end{equation}

697: The slope of ${\cal F}_{\rm K}\left(\langle M_\ast \rangle\right)$ is typical for 

698: the Kroupa initial mass function (IMF) of stars \citep{Kroupa01}. 

699: The mean disk masses are then calculated as follows

700: \begin{equation}

701: \overline{M}_{\rm \rm d,ph} = { \sum^N_{i=1} \langle M^i_{\rm d,ph} \rangle \, \,

702: {\cal F}_{\rm K}\left(\langle M^i_{\rm \ast,ph} \rangle\right) \over 

703: \sum^N_{i=1} {\cal F}_{\rm K} \left(\langle M^i_{\rm \ast,ph} \rangle\right) },

704: \label{mean}

705: \end{equation}

706: where $\langle M^i_{\rm d,ph} \rangle$ and $\langle M^i_{\rm \ast,ph} \rangle$ are the 

707: time-averaged disk and stellar masses in the $i$-th model (see Fig.~\ref{fig2})  

708: and index ``ph'' stands for Class 0 (C0), Class I (CI), or 

709: Class II (CII) stellar evolution phases. The summation is performed separately for each 

710: evolution phase. 

711: In particular, there are $N=20$ models that have Class~0 disks and $N=31$ models that have

712: Class I disks. All 32 models have Class II disks, of course. 

713: The ratio of the  normalization constants $A/B=2$ needed to evaluate 

714: equation~(\ref{mean}) is found using the continuity condition 

715: at $\langle M_\ast\rangle=0.5~M_\odot$.

716: 

717: %The slope of the initial mass function

718: %of stars is set to 1.3 in the $0.08-0.5~M_\odot$ range and 2.3 for stars more massive than 

719: %$0.5~M_\odot$ \citep{Kroupa01}. 

720: 

721: 

722: Table~\ref{table4} indicates that the mean masses slightly increase along the stellar evolution

723: sequence from $\overline{M}_{\rm d,C0}=0.09~M_\odot$ around Class~0 objects to 

724: $\overline{M}_{\rm d,CII}=0.12~M_\odot$ around Class~II objects.

725: The minimum and maximum time-averaged disk masses 

726: ($\langle M^{\rm min}_{\rm d,ph} \rangle$ and $\langle M^{\rm max}_{\rm d,ph} \rangle$, 

727: respectively) 

728: in each evolution phase are shown in the last three columns of Table~\ref{table4}. It is 

729: seen that both the Class I and Class II disks have a similar range of masses, while 

730: Class 0 disks have a somewhat narrower range. The smallest time-averaged mass of a 

731: self-gravitating disk found in our numerical simulations is $\langle M_{\rm d} \rangle = 0.02~M_\odot$.

732: 

733: 

734:  

735: 

736: 

737: 

738: \section{Masses of viscous disks}

739: \label{VD}

740: In this section we present disk masses obtained in the framework of the viscous numerical approach

741: described in \S~\ref{viscmodel}. We seek to determine the effect of viscosity,

742: and associated viscous torques, on the masses of self-consistently formed circumstellar disks.

743: Our comparison study of viscous versus self-gravitating disks have shown that

744: viscous disks may lack a sharp transition boundary between the disk and the envelope \citep{VB4}.

745: This smearing of the disk's outer physical boundary is particularly pronounced in 

746: disks of low mass and in the late Class~II phase.

747: In this situation, it is somewhat arbitrary to adopt a value of $\Sigma_{\rm tr}=0.1$~g~cm$^{-2}$

748: for the disk-to-envelope transition. Nevertheless, we have recalculated disk

749: masses with $\Sigma_{\rm tr}=0.01$~g~cm$^{-2}$ and found that this can increase 

750: masses of Class~II disks by $25\%$ at most. Class 0/I disks are unaffected by the value of 

751: $\Sigma_{\rm tr}$, since they usually have sharp outer physical boundaries.

752: 

753: 

754: \begin{figure}

755:   \resizebox{\hsize}{!}{\includegraphics{f3.eps}}

756:       \caption{The same as Figure~\ref{fig1} but obtained in the framework of the viscous 

757:       numerical approach described in \S~\ref{viscmodel}}

758:          \label{fig3}

759: \end{figure}

760: 

761: 

762: \subsection{Temporal evolution of viscous disks}

763: In order to investigate the temporal evolution of viscous disks, we consider the same 

764: four sample model cloud cores as in \S~\ref{tempevol}. Figure~\ref{fig3}

765: presents  disk masses (thick solid lines), stellar masses (thin solid lines),

766: and envelope masses (dashed lines) in model~8 (upper left), model~11 (upper right), model~16 (lower

767: left), and model~20 (lower right). The horizontal axis is time elapsed since the formation of a central

768: star. All masses are calculated relative to the corresponding 

769: initial cloud core mass $M_{\rm cl}$.

770: The vertical dotted lines mark the onset of Class I phase (left line) and Class II phase (right line).

771: 

772: 

773: The comparison of Figures~\ref{fig1} and \ref{fig3} reveals a major difference between self-gravitating

774: and viscous disks  -- the latter have 

775: considerably smaller masses in the {\it late} Class II phase than the former. 

776: For instance, model~8 in Figure~\ref{fig1} has a final ($t=5$~Myr) stellar mass 

777: $M_\ast = 0.33~M_\odot$ and disk mass   

778: $M_{\rm d}=2.5\times10^{-2}~M_\odot$, while the same model in Figure~\ref{fig3} has  

779: $M_{\rm d}=8.5\times 10^{-4}~M_\odot$.

780: Such a large contrast, however, diminishes for stars of greater mass. For instance, model~16

781: in Figure~\ref{fig1} has $M_\ast=1.57~M_\odot$ and $M_{\rm d}=0.37~M_\odot$, while the same model 

782: in Figure~\ref{fig3} has $M_{\rm d}=0.11~M_\odot$.

783: On the other hand, both the viscous and self-gravitating disks have comparable masses in the 

784: Class 0/I phases. Even the early Class II phase sees little difference between the masses 

785: of viscous and self-gravitating disks, provided that the disks are formed from 

786: molecular cloud cores of similar mass.

787: Figure~\ref{fig3} indicates that a small portion of the viscous disk material 

788: returns to the envelope in the late disk evolution. This is because we have imposed a fixed 

789: gas column density threshold for the disk-to-envelope transition, but

790: viscous disks expand in the late evolution due to the action of viscous torques, 

791: which are in fact positive in the outer disk regions \citep{VB4}. This numerical 

792: effect may slightly reduce the resulted viscous disk masses in the Class II phase.  

793: 

794: 

795: 

796: \begin{figure}

797:   \resizebox{\hsize}{!}{\includegraphics{f4.eps}}

798:       \caption{Total negative gravitational torques (solid lines) and total negative

799:       viscous torques (dashed lines) versus time elapsed since the formation of a central star

800:       in model~8 (upper left), model~11 (upper right), model~16 (lower left), and model~20 

801:       (lower right). The vertical dotted lines mark the onset of Class I~phase (left line) 

802:       and Class~II phase (right line). The torques are in units of $8.7\times 10^{40}$

803:       ~g~cm$^{2}$~s$^{-2}$.}

804:          \label{fig4}

805: \end{figure}

806: 

807: 

808: Why does the late stellar evolution phase reveal such a noticeable difference in the 

809: disk masses (and, consequently, in disk-to-star mass ratios) between viscous and self-gravitating 

810: disks and why is there little difference in the earlier phases? The reason for that can be understood

811: if we consider the temporal evolution of viscous and gravitational torques in the disk.

812: Figure~\ref{fig4} shows total negative torques due to disk self-gravity (${\cal T}_{\rm g}$,

813: solid lines) and viscosity (${\cal T}_{\rm v}$, dashed lines) in our four typical models:

814: model~8 (upper left), model~11 (upper right), model~16 (lower left), and model~20 (lower right).

815: The horizontal axis shows time elapsed since the formation of a central star.

816: The total negative torque is computed by summing up local negative torques $\tau(r,\phi)$ 

817: in each computational zone occupied by the disk. In particular, ${\cal T}_{\rm g}$ is computed as 

818: a sum of $\tau_{\rm g}(r,\phi) = -m(r,\phi) \, \partial \Phi(r,\phi) / \partial \phi$, where $m(r,\phi)$

819: and  $\Phi(r,\phi)$ are the gas mass and gravitational potential in a cell with polar coordinates $(r,\phi)$,

820: respectively. The total negative viscous torque is calculated in a similar manner by summing up 

821: all local negative viscous torques $\tau_{\rm v}(r,\phi)=r\, (\nabla \cdot \mathbf{\Pi})_\phi  

822: \, S(r,\phi)$, where $S(r,\phi)$ is the surface area occupied by a cell with polar coordinates ($r,\phi$).

823: The local torques $\tau_{g}(r,\phi)$ and $\tau_{\rm v}(r,\phi)$ are actually the local 

824: azimuthal components of the corresponding 

825: forces, acting on a fluid element with mass $m(r,\phi)$, multiplied by the arm $r$.

826: 

827: Figure~\ref{fig4} indicates that gravitational torques always prevail over viscous 

828: torques in the Class 0 and

829: Class I phases. As a result, the disk mass in these phases is determined by

830: the interplay between the mass load from an infalling envelope and the rate of mass

831: and angular momentum transport in the disk due to gravitational torques. In the Class II phase,

832: the strength of gravitational torques diminishes, partly due to a stabilizing influence of a growing

833: central star and partly due to the exhausted mass reservoir in the infalling envelope.

834: As a result, viscous torques, the strength of which shows only a mild decline with time, 

835: take a leading role in the Class II phase. They act to further decrease the disk mass when 

836: gravitational torques fail to do so. 

837: 

838: The actual time in the Class II phase when ${\cal T}_{\rm v}$ become greater than

839: ${\cal T}_{\rm g}$ is distinct for models with different initial cloud core masses (but similar 

840: ratios of rotational to gravitational acceleration $\eta$). 

841: Low-mass cloud cores form disks late in the evolution (see e.g. model~8 in Fig.~\ref{fig3}).

842: The resulting disks have low masses, merely due to the fact that most of the cloud core material 

843: has already been accreted directly onto the protostar in the pre-disk phase. As a result, 

844: gravitational instability is underdeveloped and viscous torques quickly take over the 

845: gravitational ones in the beginning of the Class II phase. 

846: Conversely, high-mass cloud cores form disks early in the evolution 

847: (see e.g. model~16 and 20 in Fig.~\ref{fig3}). The resulting disks are characterized 

848: by considerable masses. As a consequence, gravitational torques prevail over viscous torques 

849: throughout a considerable portion of the Class II phase. The interested reader 

850: is referred to \citet{VB5,VB4} for a detailed discussion on gravitational torques and

851: appearances of self-gravitating and viscous disks.

852: 

853: 

854: \subsection{Relation between disk and stellar masses}

855: In this section we derive statistical relations between time-averaged disk masses 

856: ($\langle M_{\rm d}\rangle$), time-averaged stellar masses ($\langle M_\ast \rangle$) , 

857: and time-averaged disk-to-star mass ratios ($\langle \xi \rangle$) for viscous disks 

858: in each stellar evolution phase. Time averaging is performed separately in each evolution phase 

859: over the duration of the phase.

860: Figure~\ref{fig5} shows $\langle M_{\rm d}\rangle$ (right column) and $\langle \xi \rangle$  

861: (left column) versus $\langle M_\ast \rangle$ for Class 0 objects (open triangles),

862: Class I objects (filled squares) and Class II objects (plus signs).

863: The top-to-bottom rows show the data for models with $\eta_1=1.2\times 10^{-3}$, 

864: $\eta_2=2.3\times 10^{-3}$, and $\eta_3=3.4\times 10^{-3}$, respectively. 

865: 

866: 

867: \begin{figure}

868:   \resizebox{\hsize}{!}{\includegraphics{f5.eps}}

869:       \caption{The same as Figure~\ref{fig2} only obtained in the framework of the viscous numerical

870:       approach described in \S~\ref{viscmodel}.}

871:          \label{fig5}

872: \end{figure} 

873: 

874: 

875: 

876: The comparison of Figures~\ref{fig2} and \ref{fig5} reveals a distinct feature of viscous disks

877: around objects of equal stellar mass -- Class II disk have systematically lower masses than their 

878: Class 0/I counterparts. 

879: This is in sharp contrast to self-gravitating disks studied in \S~\ref{SG} where

880: we have found that Class O/I/II objects with equal stellar mass 

881: possess self-gravitating disks of essentially similar mass (see Fig.~\ref{fig2}). 

882: The difference in masses between Class II and Class 0/I viscous disks 

883: is considerable for low-mass stars (up to a factor of 10), but it diminishes for stars 

884: of greater mass. However, we have to keep in mind that low-mass stars are statistically more

885: important than their high-mass counterparts.

886: 

887: The dependence of $\langle \xi \rangle$ on $\langle M_\ast \rangle$

888: in Figure~\ref{fig5} is similar to that of Figure~\ref{fig2} -- stars of greater mass tend 

889: to have greater disk-to-star mass ratios. However, as for the case of disk masses, 

890: Class II viscous disks have disk-to-star mass ratios that are lower than those 

891: of self-gravitating disks. For instance, the smallest $\langle \xi \rangle$ in self-gravitation disks

892: (Fig.~\ref{fig2}) is 0.05, while viscous disks have $\langle \xi \rangle$ as small as 0.01.

893: 

894: {It is interesting to compare our results with the recent work of \citet{Kratter}, who use 

895: semi-analytical models to describe the time evolution of embedded, accreting protostellar disks.

896: As in our model, they account for disk self-gravity and MRI-induced viscosity but

897: also include the effect of stellar and envelope irradiation. They obtain maximum disk-to-star

898: mass ratios in the Class~I phases of order $30\% - 40\%$, in excellent agreement with our results.

899: The disk mass in their fiducial $1~M_\odot$ model shows a temporal behaviour similar to that 

900: shown in the lower-left panel of Fig.~\ref{fig3} -- it grows in the Class~0 phase,

901: reaches a maximum in Class~I phase, and declines afterwards.  

902: They also predict that higher mass stars have relatively more massive disks and this trend

903: is expected to extend to stellar masses much greater than those considered in our work ($\la 2.0~M_\odot$).

904: 

905: 

906: We calculate the mean masses of viscous disks using equation~(\ref{mean})

907: and summarize the main properties of our viscous disks in Table~\ref{table4}.

908: It is evident that the range of time-averaged masses and the values of mean masses are quite similar

909: in both the self-gravitating and viscous disks. The latter tend to have slightly larger

910: mean disk masses in the Class 0/I phases, which is explained by the fact that viscous

911: torques tend to oppose gravitational ones in the early disk evolution \citep{VB4}.

912: The only significant difference comes from

913: Class~II viscous disks, which have a factor of two lower mean mass, 

914: $\overline{M}_{\rm CII}=0.06~M_\odot$, than that of self-gravitating disks.

915: Class II viscous disks also feature the lowest time-averaged disk mass found in our numerical

916: simulations, $\langle M^{\rm min}_{\rm d,CII} \rangle=0.004~M_\odot$. 

917: 

918: 

919: 

920: 

921: Finally, we seek to determine the relation between time-averaged disk and stellar masses 

922: in each stellar evolution phase separately and for the three phases taken altogether.

923: %for all model cloud cores of our sample, irrespective of their initial values of $\eta$.

924: Figure~\ref{fig6} shows $\langle M_{\rm d} \rangle$ versus $\langle M_\ast \rangle$

925: for our 32 model cloud cores. 

926: In particular, the open triangles, filled squares, and plus signs 

927: show the data for the Class 0, Class I, and Class II phases, respectively. 

928: It is evident that the upper-mass stars have a considerably narrower scatter in disk masses than 

929: the lower-mass stars. The least-squares best fit to {\it all} data in Figure~\ref{fig6} 

930: (solid line) yields the following relation between time-averaged stellar and disk masses 

931: (in $M_\odot$) for the three phases taken altogether

932: \begin{equation}

933: \langle M_{\rm d} \rangle = \left( 0.2\pm 0.05 \right) \, \langle M_\ast \rangle^{1.3\pm 0.15}.

934: \label{rel1}

935: \end{equation}

936: 

937: 

938: The relation between time-averaged disk and stellar masses in each stellar evolution phase

939: is, however, quite different from that expressed by equation~(\ref{rel1}).

940: The least-squares best fits performed separately for each evolution phase yield the 

941: following relations

942: \begin{eqnarray}

943: \label{rel3}

944: \langle M_{\rm d,CO} \rangle &=& \left( 0.2\pm 0.05 \right) \, \langle M_{\rm \ast,C0} \rangle^{0.7\pm 0.2}, \\

945: \langle M_{\rm d,CI} \rangle &=& \left( 0.3\pm 0.04 \right) \, \langle M_{\rm \ast,CI} \rangle^{1.3\pm0.15}, \\

946: \langle M_{\rm d,CII} \rangle &=& \left( 0.2\pm 0.08 \right) \, \langle M_{\rm \ast,CII} \rangle^{2.2\pm0.2}, 

947: \label{rel5}

948: \end{eqnarray}

949: where indices CO, CI, and CII refer to the Class~0, Class~I, and Class~II phases, respectively.

950: The dependence of disk masses on stellar masses becomes progressively steeper

951: along the sequence of stellar evolution phases. 

952: 

953: The above relations between time-averaged stellar and disk masses were derived for 

954: viscous disks, in which both viscosity and self-gravity are 

955: at work. Purely self-gravitating disks (zero viscosity) have a similar relation 

956: between time-averaged disk and stellar masses when {\it all} stellar evolution phases 

957: are considered altogether, namely equation~(\ref{rel1}).

958: However, when each phase is taken separately, the exponents are different from those derived

959: for viscous disks in equations~(\ref{rel3})-(\ref{rel5}). To avoid confusion, we provide here 

960: the relations for viscous disks only, since we believe that some sort of viscosity should 

961: be present in circumstellar disks.

962: 

963: 

964: 

965: \begin{figure}

966:   \resizebox{\hsize}{!}{\includegraphics{f6.eps}}

967:       \caption{Time-averaged disk masses versus time-averaged stellar masses for 32~models of our sample

968:       (see Tables~\ref{table1}-\ref{table3}). Open triangles, filled squares, and plus signs

969:       show the data for Class~0, Class~I, and Class~II systems, respectively. The solid line gives

970:       a least-squares best fit to all data in the plot. }

971:          \label{fig6}

972: \end{figure}

973: 

974: 

975: Observations provide conflicting evidence as to the relation between disk and stellar masses

976: in YSOs. For instance, \citet{Natta01} found a marginal correlation between disk 

977: and stellar masses, albeit with a substantial dispersion. On the other hand, \citet{Andrews}

978: claimed no correlation. The correlation in Figure~\ref{fig6} can be broken if 

979: a large population of objects occupies both the upper-left and lower-right portions of the 

980: $\langle M_{\rm d} \rangle - \langle M_\ast \rangle$ diagram. While the latter is feasible due to

981: the incompleteness of our sample of model cloud cores (we have not considered sufficiently low 

982: values of the ratio $\eta$), the former is unlikely due to the saturation of disk masses 

983: discussed in the context of Figure~\ref{fig2} and earlier considered by \citet{Adams89} and

984: \citet{Shu90} . No stars can harbour disks with masses comparable to or greater than 

985: the stellar mass! 

986: It is likely that the uncertainty in disk measurements may introduce a considerable scatter

987: to the observational data which precludes the correlation analysis.

988: 

989: 

990: 

991: 

992: \section{Discussion}

993: \label{discuss}

994: 

995: 

996: There exist a growing concern that disk masses around YSOs

997: may be systematically underestimated by conventional observational methods.

998: The simplest argument in favour of this conjecture is a high frequency of detection

999: of Jupiter-like exosolar planets. Both the core accretion and gravitational instability 

1000: models require disk masses of order $0.1~M_\odot$ for Jupiter-like planets to form, 

1001: which is about an order of magnitude 

1002: higher than median disk masses inferred by \citet{Andrews} (hereafter, AW).

1003: Another argument in favour of fairly massive disks is the fact that resolved disks 

1004: show rich non-axisymmetric structures, such as {\it multiple} spiral arms or arcs in 

1005: the case of AB Aurigae \citep{Fukagawa, Grady99} and HD 100546 \citep{Grady01}. Such structures 

1006: are difficult to sustain in low-mass disks. To remind the reader, a companion star or

1007: giant planet would  most likely trigger a two-armed spiral response in the disk of a target and 

1008: not a flocculent multi-arm structure. 

1009: 

1010: As was discussed in the Introduction, measurements of disk masses suffer from the uncertainties

1011: in dust opacities and disk radial structure, which could lead to a substantial underestimate

1012: of disk masses \citep[e.g.][]{Hartmann}. Alternatively, the disk may mask itself, especially late in

1013: the evolution, by locking a significant

1014: amount of its mass in the form of large dust grains or planetesimals which are inefficient emitters

1015: at submillimeter wavelengths \citep{Andrews}. 

1016: 

1017: Our numerical modeling supports the supposition of \citet{Hartmann} that circumstellar disks are 

1018: more massive than currently inferred from observations. The mean masses of disks 

1019: in our models range between $0.10-0.11~M_\odot$ for Class~I objects and $0.06-0.12~M_\odot$

1020: for Class II objects, well in excess of the AW's median masses for either 

1021: Class I disks ($3\times 10^{-2}~M_\odot$) or, especially, Class II disks ($3\times 10^{-3}~M_\odot$).

1022: Much the same, our Class 0 disks have mean masses $0.09-0.10~M_\odot$, while 

1023: five Class 0 disks in a sample of \citet{Brown} have a mean mass of order $0.01~M_\odot$,

1024: though the statistics in the case of Class~0 objects is inadequate to draw any firm conclusions.

1025: 

1026: A closer look at the data provided by AW reveals that our numerical modeling fails to

1027: reproduce only the lower bound on the measured disk masses but succeeds at reproducing the

1028: upper bound. For instance, figure~15 of AW

1029: indicates that Class I disk masses lie in the range $M_{\rm d,CI}=0.001-0.6~M_\odot$,

1030: while our Class I disks have masses in the range 

1031: $\langle M_{\rm d,CI} \rangle =0.02-0.55~M_\odot$. 

1032: Similarly, AW's Class II disk masses lie in the range $M_{\rm d,CII}=5\times 

1033: 10^{-4} - 0.5~M_\odot$, while our disks have masses in the range 

1034: $\langle M_{\rm d,CII} \rangle = 4\times 10^{-3}-0.7~M_\odot$.

1035: If disk masses are indeed systematically underestimated by conventional observational methods, then

1036: low-mass disks should suffer from this problem to a greater extent than do upper-mass disks.

1037: The reason for this is not well understood.

1038: 

1039: It is worth noting that our Class 0 disk have masses that are similar to those 

1040: of Class I disks. This is in agreement with the modeling of circumstellar

1041: envelope dust emission of 6 deeply embedded systems by \citet{Looney}, who have found 

1042: that Class 0 systems do not 

1043: have disks that are significantly more massive than those in Class I systems.

1044: This led them to conclude that the disk in Class 0 systems must quickly and

1045: efficiently process $\sim 1.0~M_\odot$ of material from the envelope onto the protostar.

1046: Our numerical modeling corroborates their conclusion -- Class 0/I disks develop 

1047: vigorous gravitational instability that helps keep the disk mass well below that of the protostar.

1048: 

1049: 

1050: To what extent can our numerical modeling overestimate disk masses?

1051: The reason for the overestimate may be twofold.

1052: First, Class II disk masses in Figures~\ref{fig2} and \ref{fig5} were

1053: derived from disks which have ages of order 3~Myr but real disks may be older.

1054: To test this possibility, we ran a few models for 5~Myr and compared the resulting 

1055: Class II disk masses with those obtained for 3~Myr-old disks. We found that longer ages of disks 

1056: could only cause a factor of 2 (at most) decrease in the time-averaged Class II disk masses,

1057: still not enough to reconcile our obtained disk masses with those of AW.

1058: Second, we might have not taken into account some important physical

1059: mechanisms that help reduce disk masses. 

1060: Indeed, dust disks and accretion disappear on time scales of about 6~Myr and

1061: this points to the existence of additional disk clearing mechanisms, especially 

1062: relative to the non-viscous disk models.

1063: Magnetic braking is known to be efficient 

1064: at transporting disk angular momentum directly to the external environment. 

1065: However, disks are known to have low ionization fractions and 

1066: ambipolar diffusion may strongly moderate the effect of magnetic braking.

1067: We plan to explore this mechanism in a follow-up paper.

1068: The formation of a binary system can also decrease the resulting disk masses

1069: but the fraction of binary systems is under debate and it may be too low to reconcile our

1070: model disk masses with observations \citep{Lada}. 

1071: Photoevaporation of disks due to the external ultraviolet radiation may somewhat decrease 

1072: disk masses but it operates only late in the evolution of Class II disks and is expected to

1073: have little influence on the time-averaged disk masses.

1074: Clearly,  the statistics on Class~II disks must depend on the 

1075: existence of additional disk clearing mechanism and the adopted end time in our numerical and

1076: more work is needed to resolve the cause of disparity between observationally and numerically

1077: derived disk masses.

1078: 

1079: \section{Numerical and model caveats}

1080: \label{caveats}

1081: \subsection{Gas thermodynamics}

1082: It has recently become evident that

1083: the gas thermodynamics plays an important role in regulating gravitational instability 

1084: and fragmentation in circumstellar disks \citep[e.g.][]{Pickett03,Johnson03,Rice03,Boley06,Cai08}. 

1085: While most researches agree that circumstellar disks are susceptible to gravitational instability, 

1086: particularly in the early phase of stellar evolution, 

1087: there still no consensus on whether fragmentation

1088: ever occurs and, if it does, whether any clumps that form will become bound protoplanets.

1089: For instance, \citet{Johnson03} have considered 

1090: geometrically thin disks that cool by radiatively transporting thermal energy (generated in the midplane)

1091: to the disk surface. They have found that disks are susceptible to fragmentation only if 

1092: the effective cooling time is comparable to or smaller than the dynamical time. Similar 

1093: conclusions have been made by \citet{Rice03} and \citet{Mejia04} using global SPH and grid-based

1094: numerical simulations, respectively. Fragmentation can be stabilized in the inner disks by slow cooling

1095: \citep[e.g.][]{Rafikov, Stamatellos} and in the outer disks by stellar and envelope 

1096: irradiation \citep{Matzner05,Cai08}.

1097: On the outer hand, disk fragmentation can be aided by the infall of matter from an external envelope,

1098: certainly in the early phase of disk evolution \citep{VB1,VB2,Krumholz,Kratter}. 

1099: 

1100: The accurate implementation of gas thermodynamics in numerical simulations

1101: of self-consistent disk formation and long-term evolution is a very difficult task.

1102: This is because our numerical simulations capture several physically distinct regimes 

1103: that see a substantial change with time in the chemical composition, opacities, and dust properties.

1104: Numerical techniques allowing for the accurate treatment of thermal physics 

1105: in such numerical simulations are only starting to emerge \citep[e.g.][]{Stamatellos07b}.

1106: Taking into account the complexity of gas thermodynamics in circumstellar disks, 

1107: we use the barotropic equation of state.

1108: 

1109:  It should be stressed here that circumstellar disks described by the 

1110: barotropic equation of state with $\gamma=1.4$ are susceptible to fragmentation and 

1111: formation of stable clumps in the early embedded phase of evolution. These clumps are 

1112: later driven onto the protostar and produce bursts of mass accretion that can help explain 

1113: FU Ori eruptions. This phase of protostellar accretion is known as the burst phase, 

1114: it operates in disk-star systems with mass $\ga 0.6~M_\odot$ 

1115: and it is very efficient at transporting the disk material onto the protostar \citep{VB1,VB2}. 

1116: Circumstellar disks described by $\gamma=5/3$ are hotter and less susceptible to

1117: fragmentation but the clump production is not completely suppressed \citep{VB2}.

1118: %A more realistic treatment of the thermal physics

1119: %using radiation transfer tends to produce hotter disks and create less clumps 

1120: %in the early embedded phase of disk evolution, though not completely terminating 

1121: %their production. 

1122: Recent numerical and semi-analytic modeling using a more accurate prescription for the energy

1123: balance in the disk (including radiation transfer) also predict

1124: clump formation in disks (particularly, in their outer parts) 

1125: around stars with mass equal to or more massive than one solar mass

1126: \citep{Krumholz,Mayer07,Stamatellos07,Boss08,Kratter}.

1127: 

1128: 

1129: In order to examine if a higher disk temperature can affect the accretion history and, consequently,

1130: disk masses, we stiffened the barotropic equation of state by rasing $\gamma$ from 1.4 to $5/3$.

1131: In the case of self-gravitating disks, this affects most of the disk material

1132: because most of the disk is optically thick and is characterized by surface densities 

1133: considerably larger than $\Sigma_{\rm crit}=36.2$~g~cm$^{-2}$ \citep{VB3}. On the other hand, 

1134: Class II viscous disks may have surface densities comparable to or lower than

1135: $\Sigma_{\rm crit}$, particularly in the late evolution phase \citep{VB4}. Even in this case, the increase in $\gamma$ is expected to affect

1136: a large portion of the disk material because equation~(\ref{barotropic}) 

1137: makes a gradual transition between the optically thin $\Sigma << \Sigma_{\rm crit}$ and

1138: optically thick $\Sigma>>\Sigma_{\rm crit}$ regimes. 

1139: The detailed numerical simulations are presented in \citet{VB4}, here we provide only the main 

1140: results. We find that while an increase in $\gamma$ raises the disk temperature 

1141: from 50~K (at 10~AU) to about 100~K and  makes the disk less susceptible to fragmentation 

1142: and clump formation, the amount of accreted material changes insignificantly. In fact,

1143: the mass accretion rates time-averaged over $\sim 10^4$~yr are very similar in models with $\gamma=1.4$

1144: and $\gamma=5/3$, while the instantaneous rates may be quite different.

1145: This is because a modest increase in disk temperature acts to 

1146: suppress higher order spiral modes $m > 2$, which are rather inefficient at transporting mass 

1147: radially inward and tend to produce 

1148: more fluctuations and cancellation in the net gravitational torque. At the same time,

1149: the growth of lower order modes $m \le 2$ is promoted, 

1150: which are efficient agents for mass and angular momentum redistribution. 

1151: A similar effect was reported by \citet{Cai08} for circumstellar disks heated by a strong

1152: envelope irradiation. Thus, the relative strength of lower order spiral modes increases 

1153: in hotter disks, which compensates an apparent decrease

1154: in the amount of accreted gas due to the less efficient burst phase of accretion. 

1155: 

1156: \subsection{Numerical resolution and central point mass}

1157: % The radial points are logarithmically spaced.

1158: %The innermost grid point is located at $r=5$~AU, and the size of the 

1159: %first adjacent cell lies in the range between 0.26~AU (model~21) 

1160: %and 0.36~AU (model~7). 

1161: The use of a logarithmically spaced numerical grid in the radial direction means 

1162: that the ratio $\Delta r/r$ of the cell size $\Delta r$ 

1163: to radius $r$ is 

1164: constant for a given cloud core and varies from $0.05$ (model~21) to 

1165: $0.07$ (model~7). It can be noticed from Fig.~\ref{figA1} that the aspect ratio $A=z/r$

1166: of the disk vertical scale height to radius is smaller than $\Delta r/r$ in the inner 70~AU, 

1167: which implies that a better radial (and angular) resolution is desirable in this inner region. 

1168: We have found that an increase in the resolution to $256 \times 256$ grid points makes 

1169: little influence on the accretion history and disk masses but helps save a considerable amount 

1170: of CPU time and, eventually, consider many more model cloud cores. We also point out that, due to

1171: the adopted log spacing in the radial direction, the resolution in the inner computational

1172: region is sufficient to fulfill the Truelove criterion in the disk.

1173: 

1174: In our numerical simulations the position of the central star is fixed in 

1175: the coordinate center. In reality, however, the star may wobble around the center 

1176: of mass in response to the gravity force of the disk. This

1177: can promote the growth of non-axisymmetric spiral modes (especially the m=1 mode) 

1178: in massive disks and, consequently, help limit disk masses in the early phase of stellar evolution

1179: \citep{Adams89,Shu90}. We plan to explore this potentially important mechanism in 

1180: a follow-up paper.

1181: 

1182: \subsection{The weight function}

1183: The mean disk masses listed in Table~\ref{table4} may depend on the 

1184: form of the weight function. Our adopted weight function 

1185: ${\cal F}_{\rm K}$ [see eq.~(\ref{function})] assumes that stellar masses in each

1186: stellar evolution phase are distributed according to the Kroupa IMF \citep{Kroupa01},

1187: which actually refers to the initial masses of stars that have already formed.

1188: To determine the extent to which our mean disk masses depend on the particular form

1189: of the weight function, we also consider 

1190: weight functions with a slope typical for the Salpeter IMF, 

1191: ${\cal F}_{\rm S} \left( \langle M_\ast \rangle \right)=A\, 

1192: \langle M_\ast\rangle^{-2.35}$ \citep{Salpeter}, and Miller-Scalo IMF \citep{Miller}

1193: \begin{equation}

1194: {\cal F}_{\rm MS}\left(\langle M_\ast \rangle\right) = \left\{ \begin{array}{lll} 

1195:    A \, \langle M_\ast \rangle^{-1.25} & \,\,\, 

1196:    \mbox{if $0.1~M_\odot < \langle M_\ast \rangle \le 1.0~M_\odot$ } \\ 

1197:    B \, \langle M_\ast \rangle^{-2.0} & \,\,\,

1198:    \mbox{if $1.0~M_\odot < \langle M_\ast \rangle \le 2.0~M_\odot$ } \\

1199:    C \, \langle M_\ast \rangle^{-2.3} & \,\,\,

1200:    \mbox{if $\langle M_\ast \rangle > 2.0~M_\odot$. } 

1201:     \end{array} 

1202:    \right. 

1203:    \label{MillerScalo}  

1204:  \end{equation}

1205: The resulted mean disk masses are listed in Table~\ref{table5}. It is evident that

1206: the use of different weight functions (indicated in parentheses) yields mean values that are 

1207: different from those derived from the Kroupa law by $30\%$ at most. 

1208: 

1209: 

1210: \begin{table}

1211: \begin{center}

1212: \caption{Mean disk masses with different weight functions}

1213: \label{table5}

1214: %\vspace{3 pt}

1215: \begin{tabular}{cccc}

1216: \hline

1217: \hline

1218: disk type & $ \!\!\! \overline{M}_{\rm d,C0}$ & $\!\!\! \overline{M}_{\rm d,CI}$ & $\!\!\! \overline{M}_{\rm d,CII}$  \\ 

1219: \hline

1220: self-gravitating (${\cal F}_{\rm S}$) & 0.08 & 0.08 & 0.10  \\

1221: viscous (${\cal F}_{\rm S}$) & 0.09 & 0.09 & 0.05  \\

1222: self-gravitating (${\cal F}_{\rm MS}$) & 0.09 & 0.12 & 0.15 \\

1223: viscous (${\cal F}_{\rm MS}$) & 0.10 & 0.13 & 0.08 \\

1224:  \hline

1225: \end{tabular} 

1226: \tablecomments{All disk masses are in $M_\odot$. The weight functions used to derive

1227: the mean disk masses are shown in parentheses.}

1228: \end{center}

1229: \end{table} 

1230: 

1231: 

1232: \section{Summary}

1233: \label{summary}

1234: We have considered numerically the long-term evolution of self-consistently formed 

1235: self-gravitating and viscous circumstellar disks around low-mass stars ($0.2~M_\odot \la 

1236: M_\ast\la 2.0~M_\odot$).

1237: We seek to determine time-averaged disk masses in the Class~0 ($\langle M_{\rm

1238: d, C0}\rangle$), Class~I ($\langle M_{\rm d,CI}\rangle$), 

1239: and Class~II ($\langle M_{\rm d,CII} \rangle$) stellar evolution phases. 

1240: Time averaging in each evolution phase is done over the duration of the phase.

1241: The mean disk masses ($\overline{M}_{\rm d}$) are then derived from these time-averaged masses 

1242: by weighting them over the corresponding stellar masses using a power-law function

1243: with a slope typical for the Kroupa initial mass function of star (see eq.~[\ref{mean}]). 

1244: In self-gravitating disks the radial transport of mass and angular momentum is performed 

1245: exclusively by gravitational torques, while in viscous disks both the gravitational and viscous torques

1246: are at work. Our numerical simulations yield the following results.

1247: 

1248: \begin{enumerate}

1249: \item Both the self-gravitating and viscous disks have Class~I

1250: mean masses $\overline{M}_{\rm d,CI}=0.10-0.11~M_\odot$ that are slightly more

1251: massive than those of Class~0 disks $\overline{M}_{\rm d,C0}=0.09-0.10~M_\odot$.

1252: However, viscous Class II disks have a mean mass ($\overline{M}_{\rm d,CII}=0.06~M_\odot$) that

1253: is a factor of 2 lower than that of self-gravitating Class II disks.

1254: This is explained by the fact that gravitational torques prevail through the Class 0 and Class I 

1255: phases but succumb to viscous torques through much of the Class II phase.

1256: 

1257: \item Our obtained mean disk masses are {\it larger} than

1258: those derived by AW for 153 YSOs in the Taurus-Auriga star formation region, 

1259: regardless of the physical mechanisms of mass transport in the disk.

1260: The difference is especially large for Class II disks, for which we find

1261: $\overline{M}_{\rm d,CII}=0.06-0.12~M_\odot$ but AW report median masses 

1262: of order $3\times 10^{-3}~M_\odot$.

1263: Our Class~I disks have almost a factor of 4 larger mean disk mass than those of AW.

1264: 

1265: 

1266: \item Time-averaged disk masses have a considerable scatter around mean values in each evolution phase.

1267: For instance, Class~0 and Class~I disk masses lie in the range $\langle M_{\rm d,C0} \rangle = 0.04-0.2~M_\odot$

1268: and $\langle M_{\rm d,CI} \rangle=0.02-0.55~M_\odot$, respectively, 

1269: while Class II disks have a much wider range in

1270: masses $M_{\rm d,CII}=0.004-0.7~M_\odot$.

1271: 

1272: 

1273: \item When the three stellar evolution phases are considered {\it altogether}, 

1274: we find a near-linear relation between time-averaged disk and stellar masses, 

1275: $\langle M_{\rm d} \rangle = \left( 0.2 \pm 0.05 \right) \, \langle M_\ast \rangle^{1.3\pm 0.15}$.

1276: This relation can potentially become somewhat shallower if cloud cores with very low

1277: ratios of rotational-to-gravitational acceleration are considered, but cannot be broken completely.

1278: However, when each phase is considered {\it separately}, the relation between disk and stellar 

1279: masses becomes progressively steeper along the sequence of stellar evolution phases.

1280: In particular, the corresponding relations for Class 0, Class I, and Class II objects have

1281: exponents $0.7\pm 0.2$, $1.3\pm 0.15$, and $2.2 \pm 0.2$, respectively.

1282: 

1283: \item  In each stellar evolution phase, time-averaged disk-to-star mass ratios 

1284: $\langle \xi \rangle$ tend to have greater values 

1285: for stars of greater mass. However, there is a clear saturation effect -- $\langle \xi \rangle$ 

1286: never exceed $40\%$, regardless of the stellar mass. 

1287: {\it This means that no star can harbour a disk with mass comparable to or greater than

1288: that of the star}.

1289: This saturation effect is caused by the onset of vigorous

1290: gravitational instability in circumstellar disks that form in the Class~0 or early Class~I phase

1291: \citep[see also][]{Adams89,Shu90}.

1292: 

1293: \item Only low-mass objects $M_\ast \le 0.9~M_\odot$ are expected to have Class~0 disks.

1294: Some Class~0 objects (particularly those formed from slowly rotating cloud cores)

1295: may have no disks and outflows associated with them. 

1296: %We predict that protostellar outflows do not form around these objects.

1297: 

1298: \end{enumerate}

1299: 

1300: 

1301: 

1302: \acknowledgments

1303:    The author is thankful to the referee for suggestions and comments that 

1304:    helped improve the manuscript and 

1305:    Prof. Shantanu Basu for stimulating discussions. The author

1306:    gratefully acknowledges present support 

1307:    from an ACEnet Fellowship. Numerical simulations were done 

1308:    on the Atlantic Computational  Excellence Network (ACEnet). Some of the simulations

1309:    we done on the Shared Hierarchical Academic Research Computing Network (SHARCNET)

1310:    while the author was a CITA National Fellow at the University of Western Ontario.

1311: 

1312: \appendix

1313: 

1314: \section{Disk scale height}

1315: We derive the disk vertical scale height $Z$ at each computational cell 

1316: via the equation of local vertical pressure balance 

1317: \begin{equation}

1318: \rho \, \tilde{c}_s^2 = 2\int_0^Z \rho \left( g_{z,\rm gas}+g_{z,\rm st} \right) dz,

1319: \label{eq1}

1320: \end{equation}

1321: where $\rho$ is the gas volume density, $g_{z,\rm gas}$ and $g_{z,\rm st}$ are the {\rm vertical} 

1322: gravitational accelerations due to self-gravity of a gas layer and gravitational pull of 

1323: a central star, respectively. 

1324: Assuming that $\rho$ is a slowly varying function of vertical distance $z$ between $z=0$ (midplane)

1325: and $z=Z$ (i.e. $\Sigma=2\, Z \,\rho$) and using the Gauss theorem, one can show that

1326: \begin{eqnarray}

1327: \int_0^Z \rho \, g_{z,\rm gas} \,  dz &=& {\pi \over 4} G \Sigma^2 \label{eq2} \\

1328: \int_0^Z \rho \, g_{z,\rm st} \, dz &=& {G M_\ast \rho \over r} 

1329: \left\{  1-\left[ 1+ \left({\Sigma\over 2 \rho r} \right) \right]^{-1/2} \right\},

1330: \label{eq3}

1331: \end{eqnarray}

1332: where $r$ is the radial distance and $M_\ast$ is the mass of the central star.

1333: Substituting equations~(\ref{eq2}) and (\ref{eq3}) back into equation~(\ref{eq1}) we obtain

1334: \begin{equation}

1335: \rho \, \tilde{c}_s^2 = {\pi \over 2} G \Sigma^2 + {2 G M_\ast \rho \over r} 

1336: \left\{  1-\left[ 1+ \left({\Sigma\over 2 \rho r} \right) \right]^{-1/2} \right\}.

1337: \label{height1}

1338: \end{equation}

1339: This can be solved for $\rho$ given the model's known $\tilde{c}_s^2$, $\Sigma$, and $M_\ast$

1340: using Newton-Raphson iterations.

1341: The vertical scale height is finally derived as $Z=\Sigma/(2\rho)$.

1342: 

1343: 

1344: Finally, we compare our model values of $Z$ with those  predicted 

1345: from detailed vertical structure models of irradiated accretion disks around T Tauri stars 

1346: by \citet{DAlessio}.  Their figure~1b (dashed curve) yields the following relation 

1347: between the aspect ratio $A=Z/r$ of the local scale height to radial distance and  radial distance

1348: $r$

1349: \begin{equation}

1350: A = 0.1 \left( r/100\,{\rm AU} \right)^{1/4},

1351: \label{height2}

1352: \end{equation}

1353: which is plotted by the dashed line in Figure~\ref{figA1}.

1354: The solid line in Figure~\ref{figA1} shows our obtained aspect ratio $A=Z/r$ for model~13 at t=0.3~Myr.

1355: It is evident that our $A$ is somewhat underestimated in the

1356: inner disk and overestimated in the outer disk but the disagreement between the two curves 

1357: is within a factor of unity. 

1358: 

1359: 

1360: \begin{figure}

1361: \centering

1362:   \includegraphics[width=10 cm]{f7.eps}

1363:       \caption{Aspect ratio of the disk vertical scale height to radius (Z/r)

1364:       as a function of radius (r). The thick solid line shows Z/r

1365:       calculated using equation~(\ref{height1}), while the dashed line does that

1366:       for equation~(\ref{height2}).}

1367:          \label{figA1}

1368: \end{figure}

1369: 

1370: 

1371: \section{Divergence of the viscous stress tensor}

1372: The components of $\nabla \cdot {\bl \Pi}$ in polar coordinates ($r,\phi$) are

1373: \begin{eqnarray}

1374: \left( \nabla \cdot {\bl \Pi} \right)_r &=& {1\over r} {\partial \over \partial r} r \Pi_{rr} +

1375: {1 \over r}  {\partial \over \partial \phi} \Pi_{r\phi} - {\Pi_{\phi\phi} \over r}, \\

1376: \left( \nabla \cdot {\bl \Pi} \right)_\phi &=& {\partial \over \partial r} \Pi_{\phi r}

1377: + {1\over r} {\partial \over \partial \phi} \Pi_{\phi\phi} + 2 {\Pi_{r\phi} \over r},

1378: \end{eqnarray}

1379: where we have neglected the contribution from off-diagonal components $\Pi_{rz}$ and $\Pi_{\phi z}$.

1380: The components of the viscous stress tensor $\bl \Pi$ in polar coordinates ($r,\phi$) can be found in

1381: e.g. \citet{VT}.

1382: 

1383: 

1384: \begin{thebibliography}{}

1385: 

1386: \bibitem[\protect\citeauthoryear{Adams et al.}{1989}]{Adams89}

1387: Adams, F. C., Ruden, S. P., \& Shu F. H. 1989, ApJ, 347, 959

1388: 

1389: \bibitem[\protect\citeauthoryear{Andr\'e et al.}{1993}]{Andre}

1390: Andr\'e, P., Ward-Thompson, D., \& Barsony, M. 1993, ApJ, 406, 122

1391: 

1392: \bibitem[\protect\citeauthoryear{Andrews \& Williams}{2005}]{Andrews}

1393: Andrews, S. M., \& Williams, J. P. 2005, ApJ, 631, 1134  

1394: 

1395: \bibitem[\protect\citeauthoryear{Balbus \& Hawley}{1991}]{BH}

1396: Balbus, S. A., \& Hawley, J. F. 1991, ApJ, 376, 214

1397: 

1398: %\bibitem[\protect\citeauthoryear{Basu \& Mouschovias}{1994}]{Basu94}

1399: %Basu, S., \& Mouschovias, T. Ch. 1994, ApJ, 432, 720

1400: 

1401: \bibitem[\protect\citeauthoryear{Basu}{1997}]{Basu}

1402: Basu, S. 1997, ApJ, 485, 240  

1403: 

1404: \bibitem[\protect\citeauthoryear{Boley et al.}{2006}]{Boley06}

1405: Boley, A. C., Mej\'ia, A. C., Durisen, R. H., Cai, K., Pickett, M. K., D\'Alessio, P.

1406: 2006, ApJ, 651, 517

1407: 

1408: \bibitem[\protect\citeauthoryear{Boss}{2008}]{Boss08}

1409: Boss, A. 2008, ApJ, 677, 607

1410: 

1411: \bibitem[\protect\citeauthoryear{Brown et al.}{2000}]{Brown}

1412: Brown, D. W., et al. 2000, MNRAS, 319, 154

1413: 

1414: \bibitem[\protect\citeauthoryear{Cai et al.}{2008}]{Cai08}

1415: Cai, K., Durisen, R. H., Boley, A. C., Pickett, M. K., Mej\'ia, A. C. 2008,

1416: ApJ, 673, 1138

1417: 

1418: \bibitem[\protect\citeauthoryear{Caselli et al.}{2002}]{Caselli}

1419: Caselli, P., Benson, P. J., Myers, P. C., \& Tafalla, M. 2002, 572, 238

1420: 

1421: \bibitem[\protect\citeauthoryear{D'Alessio et al.}{1999}]{DAlessio}

1422: D'Alessio, P., Calvet, N., Hartmann, L., Lizano, S., \& Cant\'o, J. 1999, ApJ, 527,

1423: 893

1424: 

1425: \bibitem[\protect\citeauthoryear{Fukagawa et al.}{2004}]{Fukagawa}

1426: Fukagawa, M., et al. 2004, ApJ, 605, L53

1427: 

1428: %\bibitem[\protect\citeauthoryear{Galli et al.}{2006}]{Galli06}

1429: %Galli, D., Lizano, s., Shu, F. H., Allen, A., 2006, ApJ, 647, 374

1430: 

1431: \bibitem[\protect\citeauthoryear{Grady et al.}{1999}]{Grady99}

1432: Grady, C. A., et al. 1999, ApJL, 523, L151

1433: 

1434: \bibitem[\protect\citeauthoryear{Grady et al.}{2001}]{Grady01}

1435: Grady, C. A., et al. 2001, AJ, 122, 3396

1436: 

1437: \bibitem[\protect\citeauthoryear{Hartmann et al.}{1998}]{Hartmann98}

1438: Hartmann, L., Calvet, N., Gullbring, E., \& D\'Alessio, P. 1998, ApJ, 495,385

1439: 

1440: \bibitem[\protect\citeauthoryear{Hartmann et al.}{2006}]{Hartmann}

1441: Hartmann, L., D'Alessio, P., Calvet, N., \& Muzerolle, J. 2006, ApJ, 648, 484 

1442: 

1443: \bibitem[\protect\citeauthoryear{Johnson \& Gammie}{2003}]{Johnson03}

1444: Johnson, B. M., \& Gammie C. F. 2003, ApJ, 597, 131

1445: 

1446: \bibitem[\protect\citeauthoryear{Klessen}{2001}]{Klessen01}

1447: Klessen, R. S. 2001, ApJ, 550, 77

1448: 

1449: \bibitem[\protect\citeauthoryear{Kratter et al.}{2008}]{Kratter}

1450: Kratter, K. M., Matzner, Ch. D., Krumholz, M. R. 2008, ApJ, 681, 375

1451: 

1452: \bibitem[\protect\citeauthoryear{Kroupa}{2001}]{Kroupa01}

1453: Kroupa, P. 2001, MNRAS, 322, 231

1454: 

1455: \bibitem[\protect\citeauthoryear{Krumholz et al.}{2007}]{Krumholz}

1456: Krumholz, M. R., Klein, R. I., \& McKee, C. F. 2007, ApJ, 656, 959

1457: 

1458: \bibitem[\protect\citeauthoryear{Lada}{1987}]{Lada87}

1459: Lada, Ch. J. 1987, in IAU Symp. 115, Star Forming Regions, editors

1460: M. Peimbert \& J. Jugaku (Dordrecht: Reidel), p. 1

1461: 

1462: \bibitem[\protect\citeauthoryear{Lada}{2006}]{Lada}

1463: Lada, Ch. J. 2006, ApJL, 640, 63L

1464: 

1465: \bibitem[\protect\citeauthoryear{Laughlin et al.}{1997}]{Laughlin97}

1466: Laughlin, G., Korchagin, V., \& Adams, F. C. 1997, ApJ, 477, 410

1467: 

1468: \bibitem[\protect\citeauthoryear{Laughlin et al.}{1998}]{Laughlin98}

1469: Laughlin, G., Korchagin, V., \& Adams, F. C. 1998, ApJ, 504, 945

1470: 

1471: %\bibitem[\protect\citeauthoryear{Lin \& Pringle}{1990}]{Lin90}

1472: %Lin, D. N. C., \&  Pringle, J. E. 1990, ApJ, 358, 515

1473: 

1474: \bibitem[\protect\citeauthoryear{Lodato \& Rice}{2004}]{Lodato04}

1475: Lodato, G.,  \& Rice, W. K. M. 2004, MNRAS, 351, 630

1476: 

1477: \bibitem[\protect\citeauthoryear{Lodato \& Rice}{2005}]{Lodato05}

1478: Lodato, G., \&  Rice, W. K. M. 2005, MNRAS, 358, 1489

1479: 

1480: \bibitem[\protect\citeauthoryear{Looney et al.}{2003}]{Looney}

1481: Looney, L. W., Mundy, L. G., \& Welch, W. J. 2003, ApJ, 592, 255

1482: 

1483: \bibitem[\protect\citeauthoryear{Lynden-Bell \& Pringle}{1974}]{Lyndenbell}

1484: Lynden-Bell, D., \& Pringle, J. E. 1974, MNRAS, 168, 603

1485: 

1486: %\bibitem[\protect\citeauthoryear{Nakano \& Nakamura}{1978}]{Nakano}

1487: %Nakano, T., Nakamura, T., 1978, PASJ, 30, 671

1488: 

1489: %\bibitem[\protect\citeauthoryear{Masunaga et al.}{1998}]{Masunaga98}

1490: %Masunaga, H., Miyama, S. M., \& Inutsuka, S. 1998, ApJ, 495, 346

1491: 

1492: \bibitem[\protect\citeauthoryear{Masunaga \& Inutsuka}{2000}]{Masunaga}

1493: Masunaga, H., \& Inutsuka, S. 2000, ApJ, 531, 350

1494: 

1495: \bibitem[\protect\citeauthoryear{Matzner \& Levin}{2005}]{Matzner05}

1496: Matzner, C. D., \& Levin, Yu. 2005, ApJ, 628, 817

1497: 

1498: \bibitem[\protect\citeauthoryear{Mayer et al.}{2007}]{Mayer07}

1499: Mayer, L., Lufkin, G., Quinn, T., \& Wadsley, J. 2007, ApJL, 661, 77

1500: 

1501: \bibitem[\protect\citeauthoryear{Mej\'ia}{2004}]{Mejia04}

1502: Mej\'ia, A. C. 2004, Ph.D. Dissertation, Indiana University

1503: 

1504: \bibitem[\protect\citeauthoryear{Miller \& Scalo}{1979}]{Miller}

1505: Miller, G. E., \& Scalo, J. M. 1979, ApJS, 41, 513

1506: 

1507: \bibitem[\protect\citeauthoryear{Natta et al.}{2001}]{Natta01}

1508: Natta, A., Grinin, V. P., Mannings, V. 2001, in Protostars and Planets IV,

1509: ed. V. Mannings, A. P. Boss, S. S. Russell (Tucson: Univ. Arizona Press),

1510: 559

1511: 

1512: \bibitem[\protect\citeauthoryear{Pickett et al.}{2003}]{Pickett03}

1513: Pickett, B. K., Mejia, A. C., Durisen, R. H., Cassen, P. M., Berry, D. K., 

1514: \& Link R. P. 2003, ApJ, 590, 1060

1515: 

1516: \bibitem[\protect\citeauthoryear{Rafikov}{2005}]{Rafikov}

1517: Rafikov, R. R. 2005, ApJ, 621, 69

1518: 

1519: \bibitem[\protect\citeauthoryear{Rice et al.}{2003}]{Rice03}

1520: Rice, W. K. M., Armitage, P. J., Bate, M. R., \& Bonnell, I. A. 2003,

1521: MNRAS, 339, 1025

1522: 

1523: \bibitem[\protect\citeauthoryear{Salpeter}{1955}]{Salpeter}

1524: Salpeter, E. 1955, ApJ, 121, 161

1525: 

1526: \bibitem[\protect\citeauthoryear{Shakura \& Sunyaev}{1973}]{SS}

1527: Shakura, N. I., \& Sunyaev, R. A. 1973, A\&A, 24, 337

1528: 

1529: \bibitem[\protect\citeauthoryear{Shu et al.}{1990}]{Shu90}

1530: Shu F. H., Tremaine S., Adams F.C., \& Ruden S. P. 1990, ApJ, 358, 495

1531: 

1532: \bibitem[\protect\citeauthoryear{Stamatellos et al.}{2007a}]{Stamatellos07}

1533: Stamatellos, D., Hubber, D. A., \& Whitworth, A. P. 2007, MNRAS, 382, L30

1534: 

1535: \bibitem[\protect\citeauthoryear{Stamatellos et al.}{2007b}]{Stamatellos07b}

1536: Stamatellos, D., Whitworth, A. P., Bisbas, T., \& Goodwin, S. 2007, A\&A, 475, 37

1537: 

1538: \bibitem[\protect\citeauthoryear{Stamatellos \& Whitworth}{2008}]{Stamatellos}

1539: Stamatellos, D., \& Whitworth, A. P. 2008, A\&A, 480, 879

1540: 

1541: 

1542: 

1543: %\bibitem[\protect\citeauthoryear{Velikhov}{1959}]{Velikhov}

1544: %Velikhov, E. P., 1959, Soviet Phys.-JETP, 36, 995 (JETP Letters 35, 1398)

1545: 

1546: \bibitem[\protect\citeauthoryear{Vorobyov \& Basu}{2005a}]{VB0}

1547: Vorobyov, E. I., \& Basu, S. 2005, MNRAS, 360, 675

1548: 

1549: \bibitem[\protect\citeauthoryear{Vorobyov \& Basu}{2005b}]{VB1}

1550: Vorobyov, E. I., \& Basu, S. 2005, ApJL, 633, L137

1551: 

1552: \bibitem[\protect\citeauthoryear{Vorobyov \& Basu}{2006}]{VB2}

1553: Vorobyov, E. I., \&  Basu, S. 2006, ApJ, 650, 956

1554: 

1555: \bibitem[\protect\citeauthoryear{Vorobyov \& Theis}{2006}]{VT}

1556: Vorobyov, E. I., \& Theis, Ch. 2006, MNRAS, 373, 197

1557: 

1558: \bibitem[\protect\citeauthoryear{Vorobyov \& Basu}{2007}]{VB3}

1559: Vorobyov, E. I., \& Basu, S. 2007, MNRAS, 381, 1009

1560: 

1561: \bibitem[\protect\citeauthoryear{Vorobyov \& Basu}{2008a}]{VB5}

1562: Vorobyov, E. I., \& Basu, S. 2008a, ApJ, 676, 139

1563: 

1564: \bibitem[\protect\citeauthoryear{Vorobyov \& Basu}{2008b}]{VB4}

1565: Vorobyov, E. I., \& Basu, S. 2008b, submitted to MNRAS

1566: 

1567: 

1568: 

1569: \end{thebibliography}

1570: 

1571: 

1572: 

1573: 

1574: 

1575: \end{document}

1576: 

1577: