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

2: %\bibpunct[, ]{}{}{;}{a}{,}{,}

3: %\usepackage{graphicx}

4: %\graphicspath{{plots/}}

5: \usepackage[dvipsnames]{color} 

6: 

7: %COMMENTS

8: %mention that param study in future paper - done

9: %discuss temperature panels - done

10: %use Q=uniform figure to jump into discussion of

11: %self-reg and swing amplification. This explains

12: %sigma profile and also leads to short discussions 

13: %of x-param and delta sigma - done

14: %r ^{-0.55} is in agreement with self similar solns - done

15: %remember to renumber all figures - done

16: %print out and read Lodato, Pickett papers

17: %rewrite analogy to convection

18: %self-regulation: heating increases Q and so does accretion onto sink - done

19: % in summary, emphasise diffusion/convection analogies and magnitude

20: % of accretion rates - these are not in abstract

21: %refer to Dullemond?

22: %talk more about swing amplification, eg in galaxies

23: 

24: % User defined command

25: \newcommand{\be}{\begin{equation}}

26: \newcommand{\ee}{\end{equation}}

27: \newcommand{\cs}{c_{\rm s}}

28: \newcommand{\etal}{et al.\ }

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

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

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

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

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

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

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

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

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

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

39: \newcommand{\mcl}{M_{\rm cl}}

40: \newcommand{\mdot}{\dot{M}}

41: \newcommand{\mdotavg}{\langle \dot{M} \rangle}

42: \newcommand{\Tavg}{\langle {\cal T} \rangle}

43: \newcommand{\mdiscavg}{\langle M_{\rm d} \rangle}

44: \newcommand{\mstaravg}{\langle M_* \rangle}

45: \newcommand{\Qcr}{Q_{\rm cr}}

46: \newcommand{\ul}{\underline{\hspace{20pt}}}

47: \def\la{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$<$}}}}

48: \def\ga{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$>$}}}}

49: \hyphenation{su-per-adia-ba-tic}

50: 

51: \begin{document}

52: 

53: \title[Protostellar disc accretion]{

54: %Accretion in nonaxisymmetric protostellar discs: the role of 

55: %gravitational torques}

56: Self-regulated gravitational accretion in protostellar discs}

57: \author[E. I. Vorobyov and S. Basu]{E. I. Vorobyov$^{1,2}$\thanks{E-mail:

58: vorobyov@astro.uwo.ca (EIV); basu@astro.uwo.ca (SB)} and Shantanu

59: Basu$^{1}$ \\

60: $^{1}$Department of Physics and Astronomy, University of Western Ontario, 

61: London, Ontario, N6A 3K7, Canada \\

62: $^{2}$ Institute of Physics, South Federal University, Stachki 194, Rostov-on-Don, Russia.}

63: 

64: \date{}

65: 

66: \maketitle

67: 

68: \label{firstpage}

69: 

70: \begin{abstract}

71: We present a numerical model for the evolution of a protostellar

72: disc that has formed self-consistently from the collapse of a molecular

73: cloud core. The global evolution of the disc is followed for several 

74: million years 

75: after its formation. The capture of a wide range of spatial

76: and temporal scales is made possible by use of the thin-disc approximation. 

77: We focus on the role of gravitational torques in transporting 

78: mass inward and angular momentum outward during different evolutionary

79: phases of a protostellar disc with disc-to-star mass ratio of 

80: order 0.1. In the early phase, when the

81: infall of matter from the surrounding envelope is substantial, mass is

82: transported inward by the gravitational torques from spiral arms that are

83: a manifestation of the envelope-induced gravitational instability in

84: the disc. In the late phase, when the

85: gas reservoir of the envelope is depleted, the distinct spiral structure 

86: is replaced by ongoing irregular nonaxisymmetric density perturbations. The 

87: amplitude of these density perturbations decreases with time, 

88: though this process is moderated 

89: by swing amplification 

90: aided by the existence of the disc's sharp

91: outer edge.

92: Our global modelling of the protostellar disc reveals that

93: there is typically a residual nonzero gravitational torque from these density

94: perturbations, i.e. their effects do not exactly cancel out in each region.

95: In particular, the net gravitational torque in the inner disc tends

96: to be negative during first several million years of the evolution, while

97: the outer disc has a net positive gravitational

98: torque. 

99: Our global model of a self-consistently formed disc shows that it is

100: also self-regulated in the late phase, so that it is near the 

101: Toomre stability limit, with a near-uniform Toomre parameter 

102: $Q\approx 1.5-2.0$. Since the disc also has near-Keplerian rotation, 

103: and comparatively weak temperature variation, it

104: maintains a near-power-law surface density profile proportional to

105: $r^{-3/2}$.

106: %The self-consistently formed discs show the gas surface density

107: %profile that is inverse proportional to $r^{-1.5}$, near-Keplerian rotation, 

108: %and self-regulation near the stability limit characterised by the Toomre

109: %parameter $Q=1.5-2.0$.

110: %In this late phase, the inward transport of mass in the inner

111: %disc appears to resemble a diffusive process. Furthermore, in analogy to 

112: %convection, the global gravitational effect on the disc is to

113: %maintain the Toomre $Q$ parameter at an approximately uniform 

114: %value. Given the power-law profiles of angular velocity

115: %(Keplerian) and relatively weak dependence on the nonuniform temperature, the

116: %result is to enforce a near-power-law surface density profile of the form

117: %$r^{-3/2}$.

118: %nonisothermality?

119: %

120: %The local gravitational

121: %torques from neighbouring density fluctuations produce local spin-ups and spin-down of gas, 

122: %leading to a predominant inward motion of mass and outward motion of specific

123: %angular momentum.

124: %Finally, the magnitude of accretion rates we calculate are in good agreement with the 

125: %observed rates on to low-mass young stellar objects.

126: %We argue that the gravitational torques alone can account for the necessary 

127: %rate of inward mass transport

128: %at distances of several tens of AU, although other mechanisms may be needed in the 

129: %innermost regions of the disc to drive the gas directly on to the protostar.

130: 

131: %The self-consistently formed discs show the gas surface density

132: %profile that is inverse proportional to $r^{-1.5}$, near-Keplerian rotation, 

133: %and self-regulation near the stability limit characterised by the Toomre

134: %parameter $Q=1.5-2.0$.

135: 

136: 

137: 

138: %observed accretion

139: %rates in young protostars and T Tauri stars

140: %

141: %numerically the collapse of molecular cloud cores starting 

142: \end{abstract}

143: 

144: \begin{keywords}

145: accretion, accretion discs --- hydrodynamics --- instabilities

146: --- ISM: clouds ---  stars: formation

147: \end{keywords} 

148: 

149: 

150: \section{Introduction}

151: 

152: The presence of finite angular momentum in a prestellar cloud core

153: provides a significant obstacle to the formation of stellar-sized objects, contributing

154: to the overall ``angular momentum problem'' of star formation. Even the

155: relatively small ($\sim 1$ km s$^{-1}$ pc$^{-1}\, \approx 10^{-14}$ rad s$^{-1}$)  

156: rotation rates measured in cloud cores \citep[e.g.][]{Goodman} imply that

157: most of the infalling matter will land on a protostellar disc rather than directly on to

158: the protostar. Indeed, discs are observed

159: or inferred around at least some class 0 and class I protostars, most T Tauri

160: stars, and even around brown dwarfs. However, the observed mass ratios

161: between the disc and central object are typically $\sim 1\%$ 

162: \citep{Andrews, Scholz1}. This implies

163: that there is an efficient mechanism to transport angular momentum outward

164: and mass inward that sets in very early during the life of the disc and even

165: while it is forming.

166: %The presence of finite and large angular momenta in precollapse cloud cores 

167: %\citep{Goodman}  implies that a protostar may form a protostellar disc

168: %in a rather early evolutionary phase. Indeed, observations of star-forming regions indicate that 

169: %at least some Class 0 and Class I protostars and most T Tauri stars

170: %are surrounded by protostellar discs. 

171: 

172: \citet{VB1,VB2} have recently modeled numerically the 

173: self-consistent collapse of a rotating molecular cloud leading to

174: the formation and evolution of a protostellar disc. During the early

175: ($< 0.5$ Myr) evolution, 

176: mass infall from the core envelope can episodically destabilise

177: the disc through gravitational instability and lead to the formation 

178: of spiral structure and dense clumps within the arms. 

179: Gravitational torques associated with the spiral arms drive the

180: clumps on to the protostar, generating mass accretion and luminosity bursts

181: comparable to those observed in FU Ori stars.

182: During this early phase, most of the protostellar mass is accreted during

183: the bursts rather than in the quiescent phase between the bursts.

184: The early burst phase terminates when the infalling envelope has lost

185: most of its gas reservoir. The protostar then enters a T Tauri phase and 

186: its subsequent evolution is characterised by a low-level accretion.

187: %that appears to be proportional to the square of the stellar mass.

188: The physical mechanism or mechanisms that drive this low-level accretion

189: are not well understood. 

190: %\citet{Balbus} have suggested that 

191: The magnetorotational instability \citep[MRI;][]{Balbus} 

192: has been suggested as the means of angular momentum and mass transport 

193: in protostellar discs. However,   

194: the thermal ionisation in most parts of T Tauri discs is

195: too low to allow sufficient magnetic coupling for the MRI to operate.

196: %discs of T Tauri stars are most likely 

197: %too cold for the MRI to operate as result of magnetic coupling 

198: %induced by thermal ionisation. 

199: This problem may

200: be mitigated by invoking nonthermal ionisation by cosmic rays

201: in the upper layers of the disc \citep{Gammie96}, or X-rays from the central 

202: star \citep{Glassgold}. \citet{Turner} have shown that the MRI-induced 

203: activity can indeed extend to the disc midplane at distances $\sim 5$ AU 

204: under some circumstances. Nevertheless, the overall ability 

205: of partial unstable

206: regions of the disc to drive effective transport throughout the disc

207: remains unresolved. Furthermore, \citet{Hartmann} have recently

208: suggested that a combination of the MRI and

209: gravitational instabilities may be necessary to explain the observed 

210: accretion rates for at least the more massive T Tauri stars.

211: 

212: Given the lingering uncertainties associated with the MRI and 

213: the additional suggestion made by \citet{Hartmann} that disc masses may have been

214: systematically underestimated (due to poorly constrained dust opacities),

215: it is tempting to consider the T Tauri accretion phase 

216: as merely the residual phase of gravitationally driven accretion.

217: %Gravitational instabilities as a plausible mechanism for

218: %the angular momentum and mass transfer in gaseous discs have been proposed by

219: %many authors. 

220: Numerical simulations of {\it isolated} protostellar discs do indicate that 

221: gravitational instabilities may work if the disc masses are sufficiently large 

222: \citep[e.g.][]{Tomley, Laughlin, Lodato05}, and

223: %or cooling times 

224: %are somewhat shorter than 

225: %the characteristic dynamical time of the disc 

226: %\citep{Gammie01, Rice03}.

227: efforts have also been made to describe the effect of such global 

228: instabilities in terms of local viscous 

229: dynamics \citep[e.g.][]{Lodato04}.

230: Numerical studies with simple prescribed cooling and heating have shown that the

231: nonaxisymmetric structure in protostellar discs washes out and density 

232: fluctuations become less than one per cent when the Toomre $Q$-parameter 

233: becomes larger than a few \citep[e.g.][]{Pickett, Lodato07}. 

234: However, it is not clear when this phase begins and gravitational 

235: instabilities cease to operate.

236: %should comment out refs for Gammie01 and Rice 03 but add Turner

237: 

238: In our opinion, the fundamental limitation of the above numerical

239: simulations is in the isolated nature of the model discs.

240: The onset of the axisymmetric phase (if any) should depend not only on the disc physics 

241: but also on the physics of the surrounding medium. The infall of gas from the

242: surrounding envelope (certainly in the early evolutionary phase) or 

243: gravitational perturbations from companions and parent nonaxisymmetric molecular 

244: cloud can drive a protostellar disc away from an axisymmetric state.  

245: A memory of initial core conditions also determines the distribution

246: of angular momentum and mass in the disc as well as its intrinsic size. 

247: Edge-effects can certainly affect the disc's ability to support 

248: persistent density fluctuations, as we emphasise in this paper.

249: As \citet{Larson} pointed out, even small density fluctuations of the order 

250: of a few per cent in a self-gravitating disc 

251: can create gravitational torques that can provide angular momentum transport

252: comparable to what is often invoked via the 

253: ad-hoc $\alpha$-viscosity mechanism. 

254: %mass and angular momentum transport on time scales of many orbital periods.

255: Observations do seem to support the existence of a pronounced nonaxisymmetry

256: in discs that are several Myr old, e.g. around AB Aurigae \citep{Fukagawa} and 

257: HD~100546 \citep{Grady}.

258: Hence, it is important for numerical models to take into account the effects of the environment

259: on the long-term evolution of protostellar discs.

260: %emphasise perturbations from outside core

261: 

262: In this paper we extend our previous numerical analysis of disc accretion

263: \citep{VB1,VB2} {\it to the T Tauri phase}. We start our

264: simulation in the prestellar phase and terminate it when the central protostar 

265: and the surrounding disc are about 3~Myr old. Such long integration times

266: are made possible by the use of the thin-disc approximation. 

267: %In our numerical simulations, the protostellar disc is not isolated but 

268: %is self-consistently interacting with the surrounding cloud core.

269: We are particularly interested in the ability of gravitational torques to drive accretion 

270: in relatively aged protostellar discs. Here, we focus on the evolution 

271: of a single model in order to present a detailed study of the internal 

272: disc dynamics. A parameter study is left for future presentation.

273: % modify or cut above since its also at beginning of Sec. 3.

274: 

275: %The paper is organised as follows. 

276: %In Section~\ref{model} we give a brief description of 

277: %the numerical code and initial conditions. The long-term evolution of the mass

278: %accretion rate is presented in Section~\ref{longterm}. Spatial profiles of important

279: %quantities in protostellar disc

280: %are presented in Section~\ref{radial}. The role of gravitational torques is discussed 

281: %in Section~\ref{torque}, and a summary is given in Section~\ref{summary}.

282: 

283: % (e.g. Larson 1985, Tomley 1991, Laughlin \& Bodenheimer 1994, Boss?, ).

284: %Linear stability analysis indicate that gaseous discs characterised by 

285: %the Toomre parameter $Q\la \sqrt{3}$ are gravitationally unstable to local perturbations

286: %\citep{Polyachenko}.

287: %These conditions are likely to be fulfilled in the early burst phase when the

288: %gas infall rate onto the disc from the surrounding envelope  is of the order of

289: %$10^{-6}~M_\odot~$yr$^{-1}$ \citep{VB1,VB2} and the disc-to-star mass ratio 

290: %is about 0.1. As the envelope depletes of its gas reservoir,

291: %it becomes more difficult for the gravitational instability to operate. But the 

292: %evolutionary time when the gravitational instability ceases to operate

293: %is not well constrained and probably depends on the disc mass. Observations indicate that 

294: %the disc-to-stellar mass ratio in most T Tauri stars is $\xi\approx 0.01$,

295: %albeit with a large scatter up to $\xi=0.1$ and more (references). 

296: %This however is a rather unreliable

297: %estimate and \citet{Hartmann} argued that the disc masses derived from dust

298: %emission might have been systematically underestimated. 

299: %Although the gravitational interaction with a 

300: %presumable companion or the parent cloud may cause the observed asymmetry, 

301: %the internal causes are not completely ruled out. 

302: 

303: 

304: \section{Model description and initial conditions}

305: \label{model}

306: 

307: We use the thin-disc approximation to compute the evolution of nonaxisymmetric rotating, 

308: gravitationally bound cloud cores. For details of the basic equations, numerical methods,

309: and numerical tests we refer the reader to \citet{VB2}. 

310: %Our method allows efficient calculation of 

311: %the long-term evolution starting from the pre-stellar phase and ending

312: %in the T Tauri phase when the central protostar is about 3~Myr old, all the while

313: %resolving a large dynamic range of spatial scales.

314: Here we briefly provide the basic equations. For simplicity, we neglect the

315: contribution of a frozen-in supercritical magnetic field 

316: \citep[accounted for in some models of][]{VB2}, which does not change the

317: main qualitative results.

318: The equations of mass and momentum transport are

319: 

320: \begin{eqnarray}

321: \label{cont}

322:  \frac{{\partial \Sigma }}{{\partial t}} & = & - \nabla _p  \cdot \left( \Sigma \bl{v}_p \right), \\ 

323: \label{mom}

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

325: \end{eqnarray}

326: where $\Sigma$ is the mass surface density, 

327: ${\cal P}$ is the vertically integrated gas pressure,

328: $\bl{v}_p = v_r \rhat + v_{\phi} \phihat$ is the velocity in the

329: disc plane, $\bl{g}_p = g_r \rhat + g_{\phi}\phihat$ is the gravitational acceleration in the disc plane,

330: and $\nabla_p = 

331: \rhat \partial/\partial r + \phihat \, r^{-1}  \partial/\partial \phi$ 

332: is the gradient along the planar coordinates of the disc.

333: The system of equations is closed with a barotropic equation

334: that makes a transition from isothermal to adiabatic evolution at a

335: critical density, i.e. 

336: \begin{equation}

337: {\cal P}=c_{\rm s}^2\Sigma

338: + c_{\rm s}^2 \Sigma_{\rm cr} \left(\Sigma \over \Sigma_{\rm cr} \right)^\gamma,

339: \label{eos}

340: \end{equation}

341: where $c_{\rm s}$ is the isothermal sound speed, $\gamma=7/5$ is the ratio of specific

342: heats, and $\Sigma_{\rm cr}=36.2$~g~cm$^{-2}$, which corresponds to a critical 

343: number density $n_{\rm cr} = 10^{11}$ cm$^{-3}$ under the assumption of vertical

344: hydrostatic equilibrium \citep[see][]{VB2}. Equation~(\ref{eos}) yields

345: a density-temperature relation that is in good agreement with the density-temperature

346: relation derived using exact spherically symmetric frequency-dependent

347: radiation transfer simulations \citep[for a detailed comparison see][]{VB2}.

348: We have also not included the effect of stellar irradiation, which can be a

349: significant heating source in the disc during the late accretion phase \citep{Chiang}. 

350: Such considerations require detailed modeling of (or assumptions about) 

351: the vertical structure (e.g. flaring) of the disc \citep[see also][]{Garaud}. 

352: For simplicity, we do not

353: keep track of the vertical structure of the disc in this study. 

354: The gravitational potential in the plane of the disc is computed as

355: \begin{eqnarray}

356:  \Phi(r,\phi,z=0,t) &=&  - G \int_0^{r_{\rm out}} r^\prime dr^\prime   

357:   \nonumber \\

358:       & &      \times  \int_0^{2\pi} 

359:                \frac{\Sigma(r^\prime,\phi^\prime) d\phi^\prime}

360:                     {\sqrt{{r^\prime}^2 + r^2 - 2 r r^\prime

361:                        \cos(\phi^\prime - \phi) }}  \, ,

362: \label{potential}

363: \end{eqnarray}

364: where $r_{\rm out}$ is the radius of the cloud core 

365: \citep[see][]{BT}.

366: 

367: Equations~(\ref{cont})--(\ref{potential}) are solved in polar coordinates

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

369: $128 \times 128$ points. The radial points are logarithmically spaced.

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

371: first adjacent cell is 0.3~AU. We initiate the accretion phase after 

372: the central surface density exceeds $\Sigma_{\rm cr}$ by introducing a ``sink cell'' at $r<5$~AU, 

373: which represents the central protostar plus some circumstellar disc material, 

374: and impose a free inflow inner boundary condition. 

375: The outer boundary (at 8000~AU) remains fixed in position and there is no radial inflow or outflow

376: allowed, i.e. the cloud has a constant mass and volume. 

377: 

378: In this paper we analyse the details of the internal disc dynamics for a single

379: model which follows the evolution of a cloud core with initial mass

380: $M_{\rm cl}=0.8~M_\odot$ and outer radius $r_{\rm out}=0.04$~pc.

381: The gas has a mean molecular mass $2.33 \, m_{\rm H}$ and is initially

382: isothermal with temperature $T=10$~K. The subsequent high-density evolution is of course

383: not isothermal due to our barotropic equation of state. 

384: The initial surface density ($\Sigma$) and angular velocity ($\Omega$) distributions 

385: are characteristic of a collapsing axisymmetric magnetically

386: supercritical core \citep{Basu}:

387: \begin{equation}

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

389: \label{dens}

390: \end{equation}

391: \begin{equation}

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

393: } -1\right].

394: \end{equation}

395: These profiles have the property that the specific angular momentum $j=\Omega r^2$ 

396: is a linear function of the enclosed mass.

397: We choose a central surface density $\Sigma_0=0.12$~g~cm$^{-2}$ 

398: (corresponding to a central number density $n_0=10^6$ cm$^{-3}$)

399: and $r_0=c_{\rm s}^2/(1.5 \, G \Sigma_0)$, so that

400: the latter is comparable to the Jeans length of an isothermal sheet. 

401: The central angular velocity is $\Omega_0=1.0$~km~s$^{-1}$~pc$^{-1}$.

402: %Table~\ref{table1} lists the parameters of our model cores. 

403: %The values of the central angular velocity $\Omega_0$ 

404: %are in general agreement with the range of 

405: %velocity gradients observed in molecular clouds \citep{Goodman}. 

406: 

407: %\begin{table}

408: %\caption{Model parameters of cloud cores.}

409: %\label{table1}

410: %\vskip 0.1cm

411: %\begin{tabular}{cccccc}

412: %\hline

413: %Model & $M_{\rm cl}$ & $\Omega_0$ & $R_{\rm out}$ & $M_{\rm d}$ & $\xi$ \\ [2 pt]

414: %\hline

415: %1  & 0.79   & 1.0  & 0.04  & $8.3\times 10^{-2}$   & $12.0\%$ \\

416: %2  & 0.33   & 1.5  & 0.02  & $ 3.4 \times 10^{-2}$ & $11.5\%$ \\

417: %3  & 1.02   & 0.9  & 0.05  & 0.11  & $12.0\%$  \\ 

418: %4  & 1.83   & 0.7  & 0.085 & 0.18   & $10.9\%$  \\

419: %\hline

420: %\end{tabular}

421: %\medskip 

422: %

423: %The cloud mass $M_{\rm cl}$ and disc mass $M_{\rm d}$ 

424: %are in $\msun$, the central

425: %angular velocities $\Omega_0$ are in km~s$^{-1}$~pc$^{-1}$, and 

426: %the cloud radii $R_{\rm out}$ are in pc. Both $M_{\rm d}$ and the  

427: %disc-to-star mass ratio $\xi$ of each model are averaged over the time

428: %period between 0.5~Myr and 3.0~Myr.

429: %

430: %\end{table}

431: 

432: 

433: 

434: 

435: \section{Long-term evolution of the mass accretion rate}

436: \label{longterm}

437: 

438: \begin{figure}

439:   \resizebox{\hsize}{!}{\includegraphics{fig1.eps}}

440:       \caption{Temporal evolution of the mass accretion rate through the inner computational boundary

441:       at $r=5$~AU (solid line) and the mass infall rate on to the disc (red

442:       line).}

443:          \label{fig1}

444: \end{figure}

445: 

446: %Our numerical simulations start in the

447: %pre-stellar phase and continue into the late accretion phase,

448: %long after the formation of a protostar and protostellar disc.

449: %The disc evolution is followed for approximately 3~Myr after the 

450: %formation of the protostar. 

451: Our simulations employing the thin-disc approximation can perform efficient calculation of 

452: the time evolution starting from the prestellar phase and ending

453: in the T Tauri phase when the central protostar is about 3~Myr old.

454: Such long-term integration coupled with resolution of a wide range of length

455: scales is currently prohibitive for three-dimensional simulations.

456: 

457: The solid line in Fig.~\ref{fig1}

458: shows the temporal evolution of the mass accretion rate 

459: $\dot{M}(t)=-2\pi r v_r \Sigma(r)$ in to the sink cell,

460: where we use the inflow velocity $v_r$ of gas at $r=5$~AU.

461: A sharp increase in the mass accretion rate at $t\approx 0.064$~Myr 

462: to a peak value $\dot{M}\approx 2.1\times 10^{-5}~M_\odot$~yr$^{-1}$ 

463: marks the formation of the central 

464: protostar. The subsequent evolution is characterised by a near-constant accretion at 

465: $\dot{M}\approx 10^{-5}~M_\odot$~yr$^{-1}$ until $t\approx 0.11$~Myr 

466: when the rarefaction wave from the outer boundary

467: approaches the protostar. This phenomenon is explained in detail by \citet{VB3}. 

468: The mass accretion rate then gradually declines until $t\approx 0.15$~Myr

469: when the first layer of gas from the infalling envelope hits a centrifugal 

470: barrier outside the sink cell and the protostellar disc begins to form. During the subsequent

471: evolution, matter initially lands on the disc rather than falling directly in to

472: the central sink. The mass accretion rate 

473: initially drops to a negligible value but rises to $\dot{M}\approx10^{-6}~M_\odot$~yr$^{-1}$

474: shortly thereafter. The enhanced accretion is 

475: attributable to newly developed

476: gravitational instability and spiral structure, induced by the 

477: continuous infall of matter from the envelope. Dense clumps form within the spiral 

478: arms and are quickly driven on to the protostar by the action of gravitational torques 

479: from the spiral arms. These episodes of clump infall produce the bursts of mass accretion 

480: seen in Fig.~\ref{fig1} between $t\approx 0.17$~Myr and $t\approx 0.25$~Myr.

481: During the bursts, $\dot{M} \approx 10^{-4}~M_\odot$~yr$^{-1}$, 

482: which is reminiscent of the FU Ori eruptions.

483: This ``burst mode'' was discussed in detail by \citet{VB1,VB2}.

484: We note that the disc mass never exceeds 

485: $\approx 0.1~M_\odot$, which is reached at 0.45~Myr.

486: Subsequently, it gradually declines to $0.077~M_\odot$

487: at 3~Myr. These numbers correspond to disc-to-star mass ratios of 0.15 and

488: 0.11, respectively.

489: 

490: In this paper, we focus on the evolution of the mass accretion rate {\it after} 

491: the bursts cease to occur. 

492: Figure~\ref{fig1} indicates that

493: $\dot{M}$ is in the range $(10^{-7}-10^{-6})~M_\odot$~yr$^{-1}$ in the 

494: quiescent phase between the bursts and then decreases gradually in the later 

495: residual accretion phase (after the burst mode has ended), reaching

496: $\dot{M}\approx 3\times 10^{-9}~M_\odot$~yr$^{-1}$ at $t=3$~Myr.

497: %It is interesting to compare the temporal evolution of the mass accretion rate 

498: %through the sink cell and the mass infall rate on

499: %the disc. As a proxy for the mass infall rate on the disc, we calculate the mass infall rate 

500: %$\dot{M}_{\rm disc}(t)=-2 \pi r v_r \Sigma(r)$ at $r=600$~AU. 

501: %In our numerical simulations, the protostellar disc is usually localised in

502: %the inner 100~AU and $\dot{M}_{\rm disc}$ should represent the infall rate on the disc with 

503: %high accuracy. We plot the temporal evolution of $\dot{M}_{\rm disc}$ in Fig.~\ref{fig1} by 

504: %the red line. 

505: It is interesting to compare $\dot{M}$ with $\dot{M}_{\rm disc}$, 

506: the mass accretion rate on to the

507: disc, calculated at $r=600$ AU. This is safely 

508: outside the centrifugal disc that is usually localised within the inner 100~AU.

509: The red line shows the evolution of $\dot{M}_{\rm disc}$. It is 

510: greater than $\dot{M}$ during the quiescent phase of the burst mode and becomes much less than

511: $\dot{M}$ after $t\approx 0.5$~Myr. {\it The bursts tend to occur during the time that}

512: $\dot{M}_{\rm disc}>\dot{M}$. In principle, this inequality can be estimated

513: observationally and used to determine if a disc is in the quiescent phase 

514: of the burst mode, i.e. in between mass accretion bursts. 

515: 

516: %It is evident that in the burst mode the mass infall rate on to the disc 

517: %$\dot{M}_{\rm disc}$ is {\it larger} than the mass accretion rate through

518: %the sink cell $\dot{M}$. 

519: %During the subsequent evolution, $\dot{M}_{\rm disc}$ quickly decreases

520: %as the infalling envelope depletes of its mass and becomes much smaller than

521: %$\dot{M}$ after $t\approx 0.5$~Myr. The bursts are only seen when $\dot{M}_{\rm disc}>\dot{M}$.

522: 

523: %In principle, this inequality can serve as an indicator for 

524: %a protostar to be in the burst mode, though between the actual mass accretion bursts.  

525: %One should keep in mind that the mass accretion on to the protostar

526: %rather than the mass accretion at 5~AU ($\dot{M}$)

527: %is usually available from observations. However, the mass accretion rate 

528: %at $5$~AU is expected to be equal or large than the true mass accretion

529: %rate on to the protostar due to a possible presence of jets and outflows

530: %in the inner 5~AU. Hence, the inequality  $\dot{M}_{\rm disc}>\dot{M}$

531: %should hold when a true mass accretion rate on to the protostar 

532: %is used instead of the mass accretion rate at 5~AU ($\dot{M}$).

533: 

534: %Both quantities are measurable and this inequality can potentially serve as an indicator for 

535: %a protostar to be in the burst mode, though between the actual mass accretion bursts.

536: 

537: 

538: %We compute the mass accretion rate $\dot{M}=-2 \pi r v_r \Sigma$ at r=5~AU,

539: %where $v_r$ is the inflow velocity of gas through the sink cell.

540: %The early evolution of the mass accretion rate ($t<0.5$~Myr) was studied in detail

541: %by Vorobyov \& Basu (\cite{VB1,VB2}). 

542: %In this paper we focus on the late evolution of $\dot{M}$ between $t=0.5$~Myr 

543: %and $t=3.0$~Myr from the formation of the protostar.

544: 

545: \section{Radial structure of protostellar discs}

546: \label{radial}

547: 

548: %\begin{figure}

549: %  \resizebox{\hsize}{!}{\includegraphics{figure2.ps}}

550: \begin{figure*}

551:  \centering

552:   \includegraphics[width=15 cm]{fig2.eps}

553:       \caption{

554: Top to bottom: azimuthally-averaged values $\overline{\Omega}$, $\overline{\Sigma}$, $\overline{T}$,

555: $\overline Q$, $\overline X$, and $\overline{\triangle \Sigma}$

556: at $t=0.2$~Myr (left column), $t=0.5$~Myr (middle column), and $t=1.5$~Myr (right column).

557:       }

558:          \label{fig2}

559: \end{figure*}

560: 

561: %\begin{figure}

562: %  \resizebox{\hsize}{!}{\includegraphics{fig3.eps}}

563: %      \caption{Same as Fig.~\ref{fig2} only at t=0.5~Myr.}

564: %         \label{fig3}

565: %\end{figure}

566: %

567: %\begin{figure}

568: %  \resizebox{\hsize}{!}{\includegraphics{fig4.eps}}

569: %      \caption{Same as Fig.~\ref{fig2} only at t=1.5~Myr.}

570: %         \label{fig4}

571: %\end{figure}

572: 

573: Insight into the phenomenon that drives accretion in the quiescent

574: phase between the bursts and in the later residual accretion phase can be obtained from

575: the radial structure of protostellar discs. Figure~\ref{fig2}

576: shows the radial profiles of the azimuthally-averaged values of 

577: $\Omega$, $\Sigma$, temperature $T$, Toomre parameter

578: $Q$, the $X$-parameter (defined later), and the relative

579: surface density perturbation $\triangle \Sigma$ 

580: at $t=0.2$~Myr (left column), $t=0.5$~Myr (middle column), and $t=1.5$~Myr (right column).

581: %The radial profiles of  $\overline{\Sigma}$, $\overline{\Omega}$, $\overline{T}$,

582: %$\overline Q$, $\overline X$, and $\overline{\triangle \Sigma}$ are obtained by averaging 

583: %densities $\Sigma$, angular velocities $\Omega$, temperatures $T$, Toomre parameters

584: %$Q$, parameters $X$, and relative density perturbations $\triangle \Sigma$

585: %over the azimuthal angle $\phi$ at a specific radius $r$.  

586: The Toomre parameter is calculated as $Q=\tilde{c}_{\rm s} \Omega 

587: /\pi G \Sigma$, where $\tilde{c}_{\rm s}=(\partial {\cal P}/ \partial

588: \Sigma)^{1/2}$ is the effective sound speed. 

589: In each computational zone ($r_i,\phi_j$) we calculate

590: %The quantity $\triangle \Sigma$ in each computational

591: %zone ($r_i,\phi_j$) is calculated as 

592: \begin{equation}

593: \triangle \Sigma(r_i,\phi_j) = {\Sigma(r_i,\phi_j) - {1\over N} \sum \limits_{j=1}^N \Sigma(r_i,\phi_j)

594: \over \Sigma(r_i,\phi_j)},

595: \end{equation}

596: where $N$ is the number of grid zones in the azimuthal direction. 

597: 

598: Several important conclusions can be drawn by analysing Fig.~\ref{fig2}, which

599: we review from top to bottom.

600: The first row shows that the protostellar disc is characterised by 

601: near-Keplerian rotation during

602: its evolution, since the stellar mass always dominates the disc mass 

603: \citep[see fig. 6 of][]{VB2}. (This justifies our use of $\Omega$ in the

604: definition of $Q$, since the epicyclic frequency $\kappa = \Omega$ for Keplerian

605: rotation.) Note the visible break in the $\Omega$ profile which marks the edge of the

606: centrifugal disc. The surface density profile (second row) is also proportional to $r^{-3/2}$ 

607: in much of the disc, with the exceptions of near the

608: inner and outer boundaries. 

609: This profile has a similar form as the minimum mass solar nebula,

610: $\Sigma = 1000\,r_{\rm AU}^{-3/2}$ g cm$^{-2}$ \citep{Weiden}.

611: The mild decrease in surface density

612: near the disc's inner boundary is caused by the

613: inflow condition on the inner computational boundary. At the same time, the

614: surface density at several tens of AU shows a sharp drop by many orders

615: of magnitude, implying that the disc has a sharp physical boundary.

616: {\it This is an important new result of our self-consistent calculation of disc formation.}

617: We emphasise that the outer computational boundary is at 8000~AU

618: and has virtually no influence on the protostellar disc dynamics or the formation

619: of its sharp edge. The disc

620: grows in radius from approximately $30$~AU at $0.2$~Myr to about $70$~AU 

621: at 3~Myr.

622: 

623: The third row shows that the disc, in contrast to the envelope, is decidedly

624: nonisothermal. As with $\Omega$ and $\Sigma$, there is a sharp drop in temperature at 

625: the disc edge. The temperature generally increases inward as the surface density

626: increases, in accordance with the barotropic equation of state.

627: Nevertheless, the variation in temperature within the disc is much less than

628: the variations in $\Omega$ and $\Sigma$, which follow power-law profiles.

629: We note that the temperature distribution in our model disc during the late

630: accretion phase is an approximate value

631: since we have not included the complex roles of stellar irradiation and radiative

632: transfer. The model of \citet{Chiang}, which accounts for heating from a 

633: central star onto a passive minimum mass solar nebula model, yields temperatures 

634: in the range 50-60 K for a disc at 10 AU, in comparison to our model temperature 

635: of 40 K at this radius at late times. More detailed models by \citet{Garaud},

636: who add viscous dissipation within the context of steady-state accretion 

637: in an $\alpha$-prescription (the resulting surface density is much shallower

638: than $r^{-3/2}$), yield even greater midplane temperatures in 

639: many cases, but generally agree with the \citet{Chiang} model for

640: an accretion rate 10$^{-8}$ \msun~yr$^{-1}$. An exact comparison of our

641: calculated temperatures with either of these models is difficult due to their

642: sensitive dependence on disc flaring, a quantity that we do not follow 

643: in this study, and due to the different surface density distribution of

644: our model discs.

645: %(resulting in a surface density profile significantly

646: %shallower than $r^{-3/2}$), find similar

647: %interior disc temperatures as \citet{Chiang} for an accretion rate of

648: %10$^{-8}$ \msun yr$^{-1}$, but greater temperatures for lower accretion rates.

649: %The solutions also depend upon the degree of disc flaring in each model 

650: %and assumptions about the nature of grains. The overall problem of calculating

651: %a self-consistent temperature within the context of our dynamical model remains

652: %beyond reach for the time being, but we are satisfied that our estimated temperatures

653: %are adequate to make the conclusions about self-regulated accretion below. 

654: 

655: The radial distribution of the azimuthally-averaged Toomre parameter $\overline{Q}$ 

656: is shown in the fourth row of Fig.~\ref{fig2}. 

657: %supports the above scenario of swing-amplified

658: %low-amplitude density perturbations. 

659: A substantial portion of the disc

660: at $t=0.2$~Myr and $t=0.5$~Myr is characterised by $\overline Q$ below a 

661: fiducial critical value

662: of $\sqrt 3$ \citep[estimated from the linear analysis of][and shown

663: by the dotted line]{Polyachenko}. 

664: %, indicating that the physical conditions are favourable 

665: %for swing amplification to operate throughout most of the disc.

666: However, at $t=1.5$~Myr $\overline Q$ becomes greater

667: than $\sqrt 3$ and the relative amplitude of density perturbations in the disc is

668: expected to diminish. In all cases, $\overline{Q}$ maintains near-uniformity in 

669: space due to a self-regulation mechanism in the disc that prevents it from falling

670: substantially below the critical value.  

671: If $Q$ falls below a critical value 

672: $\Qcr$, the gas disc becomes vigorously unstable, develops

673: a spiral structure, and may even fragment to form dense clumps within the arms.

674: Accretion on to the central sink as well as heating induced by density enhancements

675: work to restore $Q$ back toward the critical value.

676: Ultimately, the near $r^{-3/2}$ surface density profile within the disc that is 

677: illustrated in the second row is a result of this self-regulation of the Toomre

678: parameter coupled with the near-Keplerian angular velocity and relatively weak

679: temperature variation.

680: We note that our obtained distribution of the azimuthally

681: averaged Toomre parameter  is

682: qualitatively similar to that obtained by \citet{Lodato04}

683: using SPH simulations of isolated discs with a simple parametrisation for the cooling function.

684: 

685: %It is known from linear stability analysis that the Toomre parameter $Q$

686: %expresses the stability properties of rotating gaseous discs. 

687: %When $Q$ falls below a critical value 

688: %$\Qcr$, the gas disc becomes vigorously unstable, develops

689: %a spiral structure, and may even fragment to form dense clumps within the arms.

690: %When $Q > \Qcr$, the gas disc stabilises and 

691: %the spiral structure diminishes. The exact value of $\Qcr$ depends on 

692: %the rotation curve and disc geometry. For instance, \citet{Polyachenko}

693: %performed a linear stability analysis of thin gaseous discs to a local perturbation

694: %and found $Q_{\rm cr}=\sqrt{1.5}$ for a uniformly rotating fluid

695: %and $Q_{\rm cr}=\sqrt{3}$ for flat rotation (i.e. when the rotation velocity is independent of

696: %radius). Unfortunately, they did not extend

697: %their analysis to Keplerian discs. However, a mild increase in 

698: %$\Qcr$ along a sequence of rotation curves, from ``solid body'' to ``flat rotation'' to ``declining

699: %rotation'' allows us to estimate that $\Qcr$

700: %is no smaller than $\sqrt{3}$ in the case of Keplerian discs. 

701: 

702: Although $Q$ cannot drop to arbitrarily {\it low} values, what prevents the disc from 

703: quickly achieving a {\it high} $Q$ state in which

704: density fluctuations are quickly washed out due to 

705: pressure gradients or shear? 

706: We believe that the answer is swing amplification.

707: Numerical simulations indicate that the transition between stable

708: and unstable phases is usually a smooth process rather than an instantaneous switch.

709: There can exist a situation when the disc sustains low-amplitude 

710: density perturbations for a substantially long time even though the Toomre

711: parameter averaged over the whole disc is above the critical value and

712: the disc is stable {\it globally}. This phenomenon can take place if gravitational instability is 

713: powered by swing amplification. Amplification occurs when any leading spiral disturbance unwinds 

714: into a trailing one due to differential rotation \citep[][and others]{GLB,Toomre}.

715: The importance of swing amplification is in its essentially local nature -- it provides

716: transitory gravitational amplification to a local patch of gas where a leading spiral disturbance 

717: unwinds into a trailing one and the stabilising influence of shear is temporarily cancelled 

718: \citep[][p. 378]{BT}.

719: It is important to realise that swing amplification can work when the disc is stable globally 

720: but unstable {\it locally}.   

721: In this case, spiral density perturbations would be amplified locally

722: (if local conditions favour gravitational instability) but the gain would be too small

723: for gravitational instability to grow globally throughout the disc. 

724: In order for this process to work for a long time, a feedback mechanism must be present that

725: constantly feeds a gas disc with leading spiral disturbances \citep[][p. 379]{BT}. In this case,

726: a reflection of trailing spiral disturbances from the sharp outer disc edge 

727: can provide the necessary feedback mechanism.

728: 

729: It is convenient to introduce the $X$-parameter when considering the effect

730: of swing amplification.  It is defined as 

731: $X \equiv \lambda/\lambda_{\rm cr}$, where 

732: $\lambda \equiv 2\pi r/m $ is the circumferential wavelength of an $m$-armed spiral disturbance,

733: $\lambda_{\rm cr} \equiv 4\pi^2 G \Sigma/\kappa^2$ is the longest unstable wavelength 

734: in a cold disc, and $\kappa$ is the epicyclic frequency.

735: According to \citet{GLB}, the gain from the 

736: swing amplifier in a gaseous disc is greatest when $0.5\la X \la 2.5$.  

737: Generally speaking, swing amplification is most efficient when 

738: $\lambda \approx \lambda_{\rm cr}$. For $\lambda \gg \lambda_{\rm cr}$ and $\lambda \ll \lambda_{\rm cr}$, 

739: the swing amplifier is strongly moderated by local shear and gas pressure, respectively.

740: Figure~\ref{fig2} (fifth row)

741: shows the radial distribution of the azimuthally-averaged 

742: $X$-parameter in the disc for the $m=6$

743: spiral mode (solid line) and $m=10$ spiral mode (dashed line). 

744: It is evident that the favourable conditions for swing amplification are

745: present in the inner few tens of AU and the maximum gain of the swing amplifier 

746: is expected for higher order modes $m \approx 10$. Consequently, the disc is expected to develop 

747: a flocculent multi-armed spiral structure, which is indeed seen in fig.~4 of \citet{VB2}.

748: 

749: The presence of swing-amplified density perturbations is illustrated

750: in Fig.~\ref{fig2} (bottom row). The filled circles show the azimuthally-averaged 

751: relative density perturbations $\overline{|\triangle \Sigma|}$

752: at different disc radii. Note that we azimuthally average the absolute

753: values of $\triangle \Sigma$ in order to avoid a partial cancellation of

754: positive and negative relative density perturbations along the azimuth.

755: The error bars indicate the maximum positive and negative relative

756: density perturbations at a specific radius. 

757: A substantial portion of the disc at $t=0.2$~Myr is characterised by $0.1<\overline{\triangle

758: \Sigma}<0.5$. A steady increase of $\overline{\triangle

759: \Sigma}$ with radius is evident, suggesting that density perturbations are powered by the swing-amplified

760: spiral disturbances reflected off the disc outer edge. 

761: The magnitude of relative density perturbations decreases with time,

762: which agrees with a gradual stabilisation of the disc inferred from the

763: temporal evolution of $\overline{Q}$.

764: For instance,  $\overline{\triangle \Sigma}\approx 0.1$ throughout most of the disc at 

765: $t=0.5$~Myr, though local

766: relative density perturbations can fall within the range $[+0.2,-0.3]$.

767: At $t=1.5$~Myr,  $\overline{\triangle\Sigma}\approx 0.01$ everywhere except just

768: at the outer disc edge. We note that the actual densities are very low

769: there and large relative density perturbations have little  influence on the disc dynamics. 

770: 

771: %We believe this is caused by a partial transparency of the disc outer edge

772: %to the outgoing spiral disturbances. However, this effect has little influence on the disc dynamics 

773: %since the actual gas density is very small there.

774: 

775: %reflection of trailing spiral disturbances into leading ones presumably takes place. 

776: %One should remember that the density drops sharply at the disc outer edge and a large 

777: %relative density perturbation is most likely caused by a partial transparency of the outer

778: %disc edge to propagating spiral density perturbations.

779: %

780: To summarise, the nonaxisymmetry diminishes with time and is less pronounced

781: in $10^6$ yr-old discs than in $10^5$ yr-old discs. 

782: Interestingly, nonaxisymmetric structure

783: is observed in the several Myr

784: old discs around HD 100546 and AB Aurigae \citep{Grady,Fukagawa}.

785: 

786: \begin{figure}

787:   \resizebox{\hsize}{!}{\includegraphics{fig3.eps}}

788:       \caption{Surface density distribution in the protostellar disc at

789:       four evolutionary times indicated in each frame. The scale bar is

790:       in g~cm$^{-2}$. The gas with surface density below 1.0~g~cm$^{-2}$

791:       is shown with white space. The central white circle represents the

792:       protostar plus some circumstellar matter that is 

793:       unresolved in our numerical simulations. }

794:          \label{fig3}

795: \end{figure}

796: 

797: \begin{figure}

798:   \resizebox{\hsize}{!}{\includegraphics{fig4.eps}}

799:       \caption{Spatial distribution of gravitational torques at four evolutionary

800:       times corresponding to the surface density distributions in Fig.~\ref{fig3}.

801:       The positive and negative gravitational torques are shown with the

802:       shades of red and blue, respectively. The scale bars are in 

803:       units of $8.66\times 10^{40}$~g~cm$^2$~s$^{-2}$. }

804:          \label{fig4}

805: \end{figure}

806: 

807: \section{A closer look at gravitational torques}

808: \label{torque}

809: 

810: \subsection{Self-regulated gravitational accretion}

811: \label{torque1}

812: 

813: Even the slightest deviation from axial symmetry in a protostellar disc

814: produces gravitational torques. The strength of the gravitational torques 

815: contributes to the rate of angular momentum redistribution \citep{Larson}, 

816: and by implication, the mass accretion rate. In this section, we 

817: consider the effect of gravitational torques on the radial redistribution of mass

818: and angular momentum within protostellar discs.

819: 

820: %search for the causal link between the gravitational torques and

821: %mass accretion rates at different evolutionary times.

822: %{\bf XXX Nujno li eto? We define the net gravitational torque ${\cal T}$ as the sum of local gravitational

823: %torques $\tau(r,\phi)=-m_{\rm c}(r,\phi) \, \partial \Phi / \partial \phi$ in

824: %the inner  600~AU,

825: %where $m_{\rm c}(r,\phi)$ is the gas mass in a cell with the polar coordinates $(r,\phi)$.

826: %The disc is usually localised within the inner 100~AU and the rest is

827: %occupied by a vary rarefied gas. Hence, the net gravitational torque in

828: %the inner 600~AU represents with high accuracy the net gravitational torque exerted on the disc. }

829: 

830: 

831: %It is well known that negative torques drive the matter inward and the angular momentum outward

832: %and positive torques act in the opposite manner. 

833: %This effect is known from numerical simulations of galactic gaseous and stellar discs with 

834: %a well-developed (usually two-armed) spiral structure \citep[][and others]{Laughlin, Tomley,VT}. 

835: %However, one has to keep in mind that the regions characterised by negative and positive torques 

836: %are radially separated in galaxies with trailing spiral arms. More specifically,

837: %the regions inside/outside corotation are usually characterised by the

838: %negative/positive torques \citep[see][for details]{Zhang}. As a consequence,

839: %the mass is transported radially inward inside corotation and radially

840: %outward outside corotation.

841: 

842: %Protostellar discs do not show a well-developed spiral structure 

843: %even in the early evolution when the infall of matter from the envelope is large.

844: %This is due to a self-regulatory mechanism that is inherent to protostellar discs -- 

845: %if the disc becomes strongly unstable, it 

846: %resists its self-destruction by driving excess gas on to the protostar. 

847: %In other words, the disc regulates itself near the stability limit, which is clearly seen 

848: %from the $Q$-parameter distribution shown in Fig.~\ref{fig2}.

849: %We illustrate this phenomenon in Fig.~\ref{fig3},

850: %which shows the gas surface density of the protostellar

851: %disc at four evolutionary times \citep[see also fig.~4 of][]{VB2}. 

852: Figure~\ref{fig3} illustrates the gas surface density of the protostellar

853: disc at four evolutionary times \citep[see also fig.~4 in][]{VB2}.

854: It is evident that multiple spiral arms are

855: present at $t=0.2$~Myr but that they disappear by about $t=0.5$~Myr. During the

856: subsequent evolution, only low-amplitude nonaxisymmetric density 

857: inhomogeneities are present in the disc.

858: It is well known that negative torques drive matter inward and 

859: angular momentum outward, and that positive torques act in the opposite 

860: manner \citep[e.g.][and others]{Laughlin,Tomley,VT}.

861: A spatial distribution of gravitational torques in the protostellar disc

862: is shown in Fig.~\ref{fig4}, in which the positive and negative gravitational

863: torques $\tau(r,\phi)$ are plotted at four evolutionary times in shades of red 

864: and blue, respectively \citep[see also fig.~5 in][]{VB2}. The regions with near-zero 

865: $\tau(r,\phi)$ are shown 

866: with white space. 

867: %The scale bars are in the units of $8.66\times 10^{40}$~g~cm$^2$~s$^{-2}$.

868: Trailing spiral arms can be clearly seen in the distribution of

869: gravitational torques at $t=0.2$~Myr. In particular, the inner part of a spiral arm is usually 

870: characterised by negative $\tau(r,\phi)$, whereas the end of a spiral is characterised by positive

871: $\tau(r,\phi)$. 

872: 

873: \begin{figure}

874:   \resizebox{\hsize}{!}{\includegraphics{fig5.eps}}

875:       \caption{Cumulative gravitational torques ${\cal T}(r)$ as a 

876:       function of radial distance $r$ (see text for explanation)

877:       at four evolutionary times. 

878: %The integration starts at 200~AU and ends at 

879: %$r$ so that ${\cal T}(r=30~{\rm AU})$ would represent the sum of all local gravitational

880: %      torques $\tau(r,\phi)$ starting from 200~AU and ending at $r=30$~AU.

881:       Note that distance decreases to the right.}

882:          \label{fig5}

883: \end{figure}

884: 

885: 

886: \begin{figure}

887:   \resizebox{\hsize}{!}{\includegraphics{fig6.eps}}

888:       \caption{Radial profiles of gas surface density (top) and specific

889:       angular momentum (bottom) at four evolutionary times. The disc forms

890:       at $t\approx 0.15$~Myr. The inward transport of mass and the outward

891:       transport of specific angular momentum after the disc formation results

892:       in a steepening of the radial profiles of the gas surface density

893:       and specific angular momentum. }

894:          \label{fig6}

895: \end{figure}

896: 

897: The spatial distribution of gravitational torques in the later evolution ($t

898: \ga 0.5$~Myr) is particularly interesting. It shows an alternating behaviour

899: in the azimuthal direction -- local gas patches characterised by positive $\tau(r,\phi)$ 

900: are followed by patches with negative $\tau(r,\phi)$ and vice versa.  

901: We note that this pattern fluctuates with time.

902: Even though the net global effect of negative 

903: and positive torques appears (by eye) to cancel out, it is clear that the

904: net inward mass accretion exhibited by our model requires a residual

905: negative gravitational torque (due to the density inhomogeneities), 

906: at least within the inner disc.  

907: %and suggests that the net local effect of negative and positive torques 

908: %due to a density inhomogeneity does not cancel out. 

909: %%In this case, gravitational torques 

910: %due to multiple 

911: %density inhomogeneities would produce local spin-ups and spin-downs of gas and

912: %the radial transport of mass in such a disc would resemble a 

913: %radial convection (or diffusion) process when the neighbouring gas patches would constantly

914: %flip their radial positions. 

915: %We emphasise that this mechanism should not

916: %be confused with convective accretion disc models proposed by \citet{Lin},

917: %in which convection is driven in the vertical direction by the vertical

918: %temperature gradient in the disc.

919: 

920: To further investigate the net effect of 

921: local positive and negative gravitational torques when averaged over a

922: significant portion of the disc,

923: we calculate the cumulative gravitational torque

924: ${\cal T}(r)$ in a radial bin between 200~AU and $r$. For instance, ${\cal T}(r=30~{\rm

925: AU})$ would represent the sum of all local gravitational torques $\tau(r,\phi)$ starting from 200~AU and ending at $r=30$~AU.

926: The gravitational torques are expected

927: to be small at large radial distances due to both the low gas surface density 

928: and effective axial symmetry in the envelope. Figure~\ref{fig5} shows the resulting integrated 

929: gravitational torque as a function of radial distance at four evolutionary times.  

930: Clearly, it is virtually zero at large 

931: radial distances. Moving radially inward, the integrated gravitational

932: torque first becomes positive and then becomes negative. 

933: This implies that the outer parts of the protostellar disc are 

934: dominated by {\it positive} gravitational torques, whereas the inner regions

935: of the disc are dominated by {\it negative} gravitational torques. 

936: This important feature contributes to the overall 

937: mass and angular momentum redistribution within the disc.

938: 

939: The global effect of gravitational torques is illustrated in Fig.~\ref{fig6},

940: where we show the radial profiles of the gas surface density $\Sigma$ (top) and specific 

941: angular momentum $j=\Omega r^2$ (bottom) at four evolutionary times. 

942: Before disc formation at $t\approx 0.15$~Myr, the infalling envelope has 

943: a shallow radial surface density profile, essentially 

944: $\propto r^{-0.5}$, and the specific angular momentum is nearly 

945: uniform in the inner 200~AU.

946: Profiles of the form $\Sigma \propto r^{-0.5}$ and $\Omega \propto r^{-2}$

947: ($j = $ const.) are the expected self-similar profiles for 

948: the freely-falling regions inside an expansion wave but outside the

949: centrifugal disc \citep{Saigo}. 

950: After disc formation, the radial density within the disc develops

951: a much steeper profile, 

952: $\Sigma\propto r^{-1.5}$, indicating an inward transport of mass. 

953: On the other hand, the specific angular momentum in the disc 

954: is obviously transported outward, 

955: which is indicated by a developed positive slope in the radial

956: distribution of $j$. Outside the disc, the specific angular momentum 

957: remains nearly spatially uniform though growing with time. 

958: %This radial redistribution of mass and angular momentum

959: %occurs due to the action of gravitational torques in the disc.

960: 

961: 

962: %Low-amplitude spiral arms in the early phase and low-amplitude density fluctuations in the

963: %later phase produce gravitational torques, which do not cancel each other

964: %

965: %This effect is known from numerical simulations of galactic gaseous and stellar discs with 

966: %a well-developed (usually two-armed) spiral structure .

967: %The regions characterised by negative and positive torques 

968: %are radially separated in galaxies with trailing spiral arms. More specifically,

969: %the regions inside/outside corotation are usually characterised by the

970: %negative/positive torques \citep[see][for details]{Zhang}. In the case of protostellar discs,

971: %we find that such a radial separation is less obvious. 

972: 

973: 

974: 

975: %The figure suggests implies that the net effect of negative torques may be cancelled

976: %by the effect of positive torques. The situation becomes even more complicated if the

977: 

978: 

979: 

980: 

981: 

982: %\begin{figure}

983: %  \resizebox{\hsize}{!}{\includegraphics{figure5.ps}}

984: %      \caption{Mass accretion rate $\dot{M}$ versus net gravitational torque

985: %      $\cal T$ in model~1. The torques and accretion rates are plotted

986: %      every 200~yr starting from $t=0.17$~Myr and ending at $t=3.0$~Myr.

987: %      The total number of data points is 13862. The least squares fitting 

988: %      yields the power index $n=0.85$.  The gravitational torques are in

989: %       the units of $8.66\times 10^{40}$~g~cm$^2$~s$^{-2}$.}

990: %         \label{fig5}

991: %\end{figure}

992: 

993: %For the gravitationally-driven radial convective motions to transport the mass radially inward and

994: %the angular momentum radially outward, the net gravitational

995: %torque  of the disc should be negative. The filled circles in Fig.~\ref{fig5} show the 

996: %net gravitational torque $\cal T$ versus the mass accretion rate $\dot{M}$ through 

997: %the inner computational boundary. 

998: %The torques and accretion rates in model~1 are plotted every 200~yr starting from $t=0.17$~Myr 

999: %and ending at $t=3.0$~Myr. The total number of data points in Fig.~\ref{fig5} is 13862, 

1000: %which guarantees excellent statistics. When we double the number of data points by making a 

1001: %twice more frequent output, the results do not change noticeably.

1002: %One has to keep in mind that the scatter in $\cal T$ at a given $\dot{M}$ 

1003: %is substantial. Hence, the time interval should be large enough to 

1004: %ensure that the characteristic long-term temporal variations in $\cal T$ and $\dot{M}$ 

1005: %are larger than their short-term fluctuations. 

1006: %We find that $\cal T$ is indeed {\it negative}

1007: %during the disc evolution for at least 3~Myr. The largest $\cal T$ and $\dot{M}$ 

1008: %are found in the early evolutionary phase and they gradually decline during the subsequent evolution. 

1009: %Using least squares to fit the set of data points in Fig.~\ref{fig5} and

1010: %substituting the net gravitational torque with its absolute values $|\cal

1011: %T|$, we obtain 

1012: %\begin{equation}

1013: %\dot{M} = 10^{-5.2}\,|{\cal T}|^{0.85\pm ?}.

1014: %\label{relation1}

1015: %\end{equation}

1016: %

1017: %

1018: %\begin{figure}

1019: %  \resizebox{\hsize}{!}{\includegraphics{figure5a.ps}}

1020: %      \caption{Mass accretion rate $\dot{M}$ versus net gravitational torque

1021: %      $\cal T$ in three time intervals {(\bf a)} $(0.17-1.0)$~Myr, {(\bf

1022: %      b)} $1.0-2.0$~Myr, and {(\bf c)} $2.0-3.0$~Myr. Solid lines show

1023: %      the least squares fitting and $m$ gives the corresponding exponents. 

1024: %      The gravitational torques are in the units of $8.66\times 10^{40}$~g~cm$^2$~s$^{-2}$. }

1025: %         \label{fig6}

1026: %\end{figure}

1027: 

1028: %Equation~(\ref{relation1}) shows that the mass accretion rate depends near-linearly 

1029: %on the net gravitational torque with the exponent $n=0.85\pm?$. The universal

1030: %character of this near-linear dependence in protostellar discs around low-mass

1031: %stars can be established by computing the

1032: %models with different initial cloud core masses and different disc masses.

1033: %Figure~\ref{fig6} shows the mass accretion rates versus net gravitational

1034: %torques in model~2 (top), model~3 (middle), and model~4 (bottom). 

1035: %The torques and accretion rates are plotted every 200~yr starting from $t=0.13$~Myr

1036: %(model~2), $t=0.2$~Myr (model~3), and $t=0.3$~Myr (model~4)  and ending at $t=2.4$~Myr.

1037: %The parameters of the models are given in Table~\ref{table1}.

1038: %The initial cloud core masses are chosen to represent the spectrum of final stellar masses 

1039: %seen in T Tauri stars. The exponent $n$ of the best square fitting is shown

1040: %in each frame of Fig.~\ref{fig6}. It is evident that the near-linear relation

1041: %between $\dot{M}$ and ${\cal T}$ holds across a wide range of initial

1042: %cloud core masses starting from $M_{\rm cl}=0.33~M_\odot$ (model~2) and

1043: %ending at $M_{\rm cl}=1.8~M_\odot$ (model~4). The net gravitational torque

1044: %${\cal T}$ is maximal in the early phase of disc evolution and gradually declines at

1045: %later times, remaining negative during approximately 3~Myr. This must represents a fundamental 

1046: %property of protostellar discs. The collapse of cloud cores almost certainly

1047: %results in most of the material initially landing at large disc radii \citep{Hartmann}.

1048: %This material should be processed through the disc and onto the central

1049: %protostar, which can only be possible if the net gravitational torque in

1050: %the inner several hundred AU is negative.

1051: 

1052: 

1053: \subsection{Diffusive nature of the accretion process}

1054: \label{torque2}

1055: 

1056: A diffusive nature of the radial mass transport in

1057: the late evolution of the disc can be visualised by considering the temporal evolution of a radially narrow 

1058: gas annulus located initially at some distance from the inner disc boundary. 

1059: Any mechanism

1060: of radial mass and angular momentum transport will change the radial 

1061: position and shape of the annulus during the subsequent evolution. 

1062: %If such a mechanism is indeed present, 

1063: %then the annulus would change its radial position and/or shape in response to the

1064: %torques. Thus, the temporal evolution of the annulus can give us insight into 

1065: %the nature of the radial transport mechanism in a protostellar disc. 

1066: 

1067: To perform this test, we

1068: follow the evolution of a radial gas annulus initially located between 20~AU and 25~AU.

1069: We solve a separate continuity equation for the annulus starting at $t=0.5$~Myr,

1070: when the protostellar disc has reached a radial size of approximately 50~AU. 

1071: The initial gas surface density of the annulus is shown in the upper panel

1072: of Fig.~\ref{fig7} with the dashed line. It equals the corresponding gas

1073: surface density in the disc and is set to a negligibly small value

1074: elsewhere. The resulting gas surface

1075: density distribution of the annulus at $t=0.75$~Myr is shown in the bottom

1076: panel of Fig.~\ref{fig7} with the

1077: dashed lines. For comparison, the solid line shows the gas surface density in the

1078: disc at the same evolutionary time. Clearly, the annulus

1079: spreads out rather than moves homologously in the radial direction. 

1080: This behaviour is in agreement with a diffusive nature

1081: of the radial mass transport in a protostellar disc.

1082: Because the net gravitational torque is negative, the annulus

1083: spreads out predominantly  inward, resulting in a radially declining surface density

1084: profile at $t=0.75$~Myr. To quantify this effect, we find the percentile of 

1085: the initial mass of the annulus ($1.04\times 10^{-2}~M_\odot$) that was transported 

1086: inwards and outwards of its initial location between 20~AU and 25~AU. 

1087: At $t=0.75$~Myr, the mass inside 20~AU is $5.2\times10^{-3}~M_\odot$ or 50 per cent 

1088: of the initial annulus mass and the mass outside 25~AU is $3.1\times 10^{-3}~M_\odot$ or

1089: 30 per cent of the initial annulus mass. These percentiles show an even 

1090: greater contrast during the subsequent evolution at $t>0.75$~Myr.

1091: %, the local action of which is not cancelled. 

1092: %We emphasise a radially declining shape of the spread-out annulus at $t=0.75$~Myr. 

1093: %This suggests a predominantly inward transport of mass, 

1094: %in agreement with a negative sign of the net gravitational torque in the disc.

1095: Another intriguing result is that the shape of the gas surface 

1096: density profile in the

1097: spread-out annulus (dashed line) closely resembles that of the disc (solid line) 

1098: after $t=0.75$~Myr. 

1099: %It suggests that the $r^{-1.5}$ radial surface density profile 

1100: %obtained in our numerical simulations (see Fig.~\ref{fig2}) may be a result of superposition

1101: %of individual radial annuli that are diffusively spread out by the action of positive

1102: %and negative gravitational torques. 

1103: %\textcolor{blue}{XXX. I am not sure that

1104: %the last sentence should be kept in the text. I am also not sure that the

1105: %discussion related to Figure 8 should be in the paper. It seems that

1106: %the paper has enough results and Figure 8 can be omitted. XXX}. 

1107: 

1108: 

1109: \begin{figure}

1110:   \resizebox{\hsize}{!}{\includegraphics{fig7.eps}}

1111:       \caption{Gravitationally-driven diffusion of a radial annulus.

1112:       Top panel: 

1113:       the disc surface density distribution (solid line) and 

1114:       the surface density in an annular region (dashed line)

1115:       at $t=0.5$~Myr. Bottom panel:

1116:       the disc surface density (solid line) and the surface density of mass previously

1117:       in the annulus, at $t=0.75$~Myr. 

1118:       The mass originally in the annulus has spread out rather than 

1119:       been translated homologously, and has a radially

1120:       declining surface density profile. This indicates a 

1121:       predominantly inward mass transport.

1122:       }

1123:          \label{fig7}

1124: \end{figure}

1125: 

1126: 

1127: %The presence of a flocculent multi-armed spiral structure in the early disc evolution

1128: %and its absence in the later evolution suggests that the radial mass transport mechanism 

1129: %may differ in the $10^5$~yr-old  and  $10^6$~yr-old discs. This conjecture is also supported 

1130: %a distinct difference between the spatial distributions of $\tau(r,\phi)$ seen in the early 

1131: %and late ...

1132: %It is important to know when the transport mechanism in the disc 

1133: 

1134: %It is illustrative to compare the mass accretion rate through the inner absorbing (inflow)

1135: %computational boundary at 5~AU with the net mass flow at different radii in the disc. Figure~\ref{fig8}

1136: %shows the temporal evolution the mass accretion rate $\dot{M}$ (black lines), the net mass fluxes

1137: %at $r=10$~AU and $r=15$~AU (red lines), and the mass infall rate on to the disc

1138: %$\dot{M}_{\rm disc}$ (blue lines). The net mass flux at $r=10$~AU is 

1139: %averaged over $10^4$~yr to wash 

1140: %out short-term fluctuations in the radial flow of gas. The net mass flux at $r=15$~AU 

1141: %is calculated in a similar manner. Only positive values of $\dot{M}_{\rm 10AU}$ and $\dot{M}_{\rm 15AU}$  

1142: %are plotted on the log scale. It is evident that the temporal behaviour of 

1143: %$\dot{M}_{\rm 10AU}$ and $\dot{M}_{\rm 15AU}$ is similar to that of the mass accretion 

1144: %rate during the early evolution, $t\la 0.6$~Myr. Of course, the mass accretion bursts are not resolved

1145: %in the net mass fluxes due to averaging over a much longer time than the duration of the

1146: %bursts ($\sim 100$~yr) but a general order-of-magnitude agreement is encouraging. It shows that 

1147: %in the early evolutionary phase ($t\la 0.6$~Myr) the {\it inner disc} is characterised by 

1148: %a dominant inflow radial motion of gas, which is probably driven by 

1149: %the negative gravitational torques from flocculent spiral arms (see Fig.~\ref{fig3}

1150: %and Fig.~\ref{fig4}).

1151: %

1152: %In the later evolution after $t\approx 0.6$~Myr, the temporal behaviour of the 

1153: %net mass fluxes changes qualitatively -- $\dot{M}_{\rm 10AU}$ and $\dot{M}_{\rm 15AU}$ 

1154: %can become both positive and negative.

1155: %This implies that, when averaged over $10^4$~yr, the gas flow in the inner 15~AU 

1156: %can be directed both radially inward and outward. These alternating flows of gas  

1157: %are most likely caused by a convective (diffusive) nature of the radial mass transport in the

1158: %late disc evolution. The inner absorbing computational boundary cannot capture this phenomenon

1159: %and the mass accretion rate $\dot{M}$ through this boundary is always positive.

1160: %Nevertheless, $\dot{M}$ and both $\dot{M}_{\rm 10AU}$ and $\dot{M}_{\rm 15AU}$ (when positive)

1161: %show a general agreement in the late evolution between 0.6~Myr and 3.0~Myr.

1162: %To quantify this agreement, we calculated the mean values of $\dot{M}$ and $\dot{M}_{\rm 10AU}$

1163: %over the time interval $(0.6-3.0)$~Myr. We find that

1164: %the mean mass accretion rate through the inner computational

1165: %boundary is $\langle \dot{M}\rangle =7.8\times 10^{-9}~M_\odot$~yr$^{-1}$, whereas 

1166: %the mean mass flux at $r=10$~AU is $\langle \dot{M}_{\rm 10AU}\rangle=2.5\times 10^{-9}~M_\odot$~yr$^{-1}$.

1167: %This indicates that the mass accretion rate through the inner computational boundary 

1168: %in the late disc evolution may be overestimated by a factor of three.

1169: %

1170: %Figure~\ref{fig8} indicates that a qualitative change in the temporal 

1171: %behaviour of $\dot{M}_{\rm 10AU}$ and $\dot{M}_{\rm 15AU}$ (red lines) 

1172: %takes place at approximately 0.6~Myr. 

1173: %%The net mass fluxes show a dominant inflow radial 

1174: %%motion of gas at $t\la 0.6$~Myr, and an alternating inflow-outflow radial motion  after 0.6~Myr.

1175: %It is interesting to note that at approximately the same time the mass infall rate on the

1176: %disc $\dot{M}_{\rm disc}$ (blue lines) drops below the mass accretion rate $\dot{M}$ on 

1177: %to the protostar (black lines) and continues to decrease afterward. This suggests that 

1178: %the infall of mass on the disc  from 

1179: %the surrounding envelope may act as an agent that sustains the disc near the stability limit.

1180: %As the envelope is depleted of matter, a weak spiral structure in the disc 

1181: %transforms into 

1182: %low-amplitude density fluctuations (see Fig.~\ref{fig3}) and the radial mass transport changes its character 

1183: %from a dominant inflow radial motion to inflow-outflow radial pulsations. 

1184: % 

1185: %\begin{figure}

1186: %  \resizebox{\hsize}{!}{\includegraphics{fig8.eps}}

1187: %      \caption{Temporal evolution of the mass accretion rate at various locations.

1188: %      (a) $\mdot$ at the inner computational boundary $r=5$~AU (black line) and at 

1189: %       $r=10$~AU (red line).

1190: %      (b) $\mdot$ at $r=5$~AU (black line) and at $r=15$~AU (red line). 

1191: %      The blue line shows the mass accretion rate on to the disc at $r=600$~AU

1192: %      in both panels.  Only positive values of the mass fluxes are plotted on the log scale.}

1193: %         \label{fig8}

1194: %\end{figure}

1195: 

1196: 

1197: %\section{Gravitational torques versus mass accretion rates across the T Tauri mass spectrum}

1198: %

1199: %So far we have considered model~1 with the initial cloud core mass 

1200: %$M_{\rm cl}=0.8~M_\odot$~yr$^{-1}$. The mean disc-to-star mass ratio

1201: %in the T~Tauri phase (between 0.5~Myr and 3.0~Myr) is $\xi 0.12$.

1202: %Measurements of the disc-to-star mass ratio in T Tauri stars show

1203: %that this quantity usually lies near 0.01, albeit with a large scatter (references).

1204: %Some objects indeed have $\xi \ga 0.1$. However, the accuracy of these measurements 

1205: %totally relies upon poorly constrained dust opacities. As \citet{Hartmann} recently 

1206: %emphasised, the disc masses may be considerably underestimated and a higher disc-to-mass

1207: %ratio of $\xi\approx 0.1$ among T Tauri stars cannot be ruled out.

1208: %...

1209: %

1210: %\begin{figure}

1211: %  \resizebox{\hsize}{!}{\includegraphics{figure8.ps}}

1212: %      \caption{}

1213: %         \label{fig8}

1214: %\end{figure}

1215: %

1216: 

1217: \section{Summary and Discussion}

1218: 

1219: Our model of the self-consistent formation of a protostellar 

1220: disc from the collapse of its parent cloud core and the subsequent

1221: long-term evolution of the disc reveals that it settles into

1222: a self-regulated state. In this state, the gravitational torques

1223: are responsible for inward mass transport 

1224: at a rate that is in agreement with typical accretion rates 

1225: inferred in T Tauri star discs \citep{Hartmann98b}. 

1226: Persistent nonaxisymmetric density perturbations are the key

1227: ingredient that leads to the gravitational torques. 

1228: The amplitude of these nonaxisymmetric density perturbations

1229: decreases with time but the decline is moderated by the 

1230: action of swing amplification, aided by the presence of a

1231: self-consistently-formed sharp edge at the interface between the

1232: protostellar disc and core envelope. The disc self-regulation

1233: by gravitational torques maintains a near-uniform value of the

1234: azimuthally-averaged Toomre $Q$ parameter. The mass accretion

1235: also maintains the disc-to-star mass ratio to be $\approx 0.1$,

1236: and the rotation rate within the disc is approximately Keplerian.

1237: Since the temperature variation in the disc is much smaller in

1238: magnitude than the power-law variation in $\Omega$, the 

1239: self-regulated state with near-uniform $Q$ implies that

1240: the surface density $\Sigma \propto r^{-3/2}$.

1241: 

1242: The near uniformity of $Q$ may be understood using an analogy

1243: to convection in stellar envelopes. A highly superadiabatic

1244: temperature gradient cannot be maintained in a stellar envelope,

1245: since convection will set in and bring the

1246: temperature gradient back to a near-adiabatic value. Similarly,

1247: gravitational torques work to maintain $Q$ at essentially the 

1248: stability limit. However, it must be emphasised that {\it

1249: our model discs are not turbulent}. There are  

1250: density perturbations in the disc, but they are not fed in 

1251: globally from the largest scales of the disc and are indeed declining in

1252: amplitude over time. The swing amplification mechanism does

1253: sustain the structure for a substantially long time, but is 

1254: driven by transient local gravitational instabilities. 

1255: We do not believe that the labels ``turbulent'' or even 

1256: ``gravitoturbulent'' are justified: rather we refer to this mode

1257: as ``self-regulated gravitational accretion''.

1258: 

1259: The $\Sigma \propto r^{-3/2}$ profile is in agreement

1260: with estimates of the minimum mass solar nebula \citep{Weiden}, made

1261: by adding the solar abundance of light elements to each planet and

1262: spreading the masses through zones surrounding the planetary orbits.

1263: The disc-to-star mass ratio in our model remains greater than 

1264: that inferred for the solar nebula or for typical external star-disc

1265: systems. However, \citet{Hartmann}

1266: have recently pointed out that disc masses based on dust emission

1267: may be systematically underestimated. In any case, other transport

1268: mechanisms, primarily associated with magnetic fields, may indeed

1269: operate in real discs, and may also be more important in the inner

1270: 5 AU that we do not model. 

1271: Since the level of magnetic coupling in the disc also remains 

1272: uncertain, our model provides a very clear 

1273: physical baseline for the global behaviour of discs before 

1274: accounting for other more poorly-understood effects.

1275: 

1276: %Our long-term global modelling of discs are accomplished efficiently

1277: %using the thin-disc approximation. For example, the simulation 

1278: %presented in this paper took XXX hours of cpu time running on a

1279: %single XXX Ghz Opteron processor. This points the way to running

1280: %a large suite of models for future comparison with specific observational

1281: %results. Semi-analytic modelling of 

1282:  

1283: %describe self-regulated state - 

1284: %emphasise diffusion/convection analogies and magnitude

1285: % of accretion rates - these are not in abstract

1286: %refer to Weidenschilling paper

1287: %mention sharp edge to disc

1288: %back to disc mass, may be underestimated

1289: %or else these results provide baseline before including other effects

1290: %i e other sources of transport could be present

1291: %discuss inner disc - possibility of MRI

1292: %emphasise disc is nonisothermal

1293: %global modeling here

1294: %disc is not turbulent

1295: %Dullemond paper?

1296: \label{summary}

1297: 

1298: 

1299: 

1300: \section*{Acknowledgments} 

1301: We thank the referee and Wolfgang Dapp for helpful comments on 

1302: the manuscript, and 

1303: Martin Houde and the SHARCNET consortium for access to 

1304: computational facilities.

1305: This work was supported by the Natural Sciences and Engineering

1306: Research Council of Canada, RFBR grant 06-02-16819-a, South Federal

1307: University grant 05/6-30, and Federal Agency of Education 

1308: (project code RNP 2.1.1.3483).  EIV gratefully acknowledges support 

1309: from a CITA National Fellowship 

1310:  

1311: 

1312: 

1313: \begin{thebibliography}{}

1314: 

1315: %\bibitem[\protect\citeauthoryear{Alexander \& Armitage}{2006}]{AA}

1316: %Alexander R. D., Armitage, P. J., 2006, ApJ, 639, L83

1317: 

1318: %\bibitem[\protect\citeauthoryear{Armitage et al.}{2001}]{Armitage} 

1319: %Armitage, P. J., Livio, M., Pringle, J. E., 2001, MNRAS, 324, 705

1320: 

1321: \bibitem[\protect\citeauthoryear{Andrews \& Williams}{2005}]{Andrews}

1322: Andrews, S. M., Williams, J. P., 2005, ApJ, 631, 1134 

1323: 

1324: \bibitem[\protect\citeauthoryear{Balbus \& Hawley}{1998}]{Balbus}

1325: Balbus, S. A., Hawley, J. F., 1998, Rev. Mod. Phys., 70, 1

1326: 

1327: \bibitem[\protect\citeauthoryear{Basu}{1997}]{Basu}

1328: Basu, S., 1997, ApJ, 485, 240  

1329: 

1330: \bibitem[Binney \& Tremaine(1987)]{BT}

1331: Binney, J., \& Tremaine, S., 1987, Galactic Dynamics. 

1332: Princeton Univ. Press, Princeton

1333: 

1334: \bibitem[\protect\citeauthoryear{Chiang \& Goldreich}{1997}]{Chiang}

1335: Chiang, E. I., Goldreich, P., 1997, ApJ, 490, 368

1336: 

1337: %\bibitem[\protect\citeauthoryear{Clarke \& Pringle}{2006}]{Clarke}

1338: %Clarke, C. J., Pringle, J. E., 2006, MNRAS, 

1339: %370, L10

1340: 

1341: %\bibitem[\protect\citeauthoryear{Dullemond et al.}{2006}]{Dullemond} 

1342: %Dullemond, C. P., Natta, A., Testi, L., 2006, ApJ, 645, L69

1343: 

1344: \bibitem[\protect\citeauthoryear{Fukagawa et al.}{2004}]{Fukagawa}

1345: Fukagawa, M., Hayashi, M., Tamura, M., Itoh, Y. et al., 2004, ApJ, 605, L53

1346: 

1347: \bibitem[\protect\citeauthoryear{Gammie}{1996}]{Gammie96}

1348: Gammie, C. F., 1996, ApJ, 457, 355

1349: 

1350: %\bibitem[\protect\citeauthoryear{Gammie}{2001}]{Gammie01}

1351: %\ul., 2001, ApJ, 553, 174

1352: 

1353: \bibitem[\protect\citeauthoryear{Garaud \& Lin}{2007}]{Garaud}

1354: Garaud, P., Lin, D. N. C., 2007, ApJ, 654, 606

1355: 

1356: \bibitem[\protect\citeauthoryear{Glassgold et al.}{1997}]{Glassgold}

1357: Glassgold, A. E., Najita, J., Igea, J., 1997, ApJ, 480, 344

1358: 

1359: \bibitem[\protect\citeauthoryear{Goldreich \& Lynden-Bell}{1965}]{GLB}

1360: Goldreich, P., Lynden-Bell, D., 1965, MNRAS, 130, 125

1361: 

1362: \bibitem[\protect\citeauthoryear{Goodman et al.}{1993}]{Goodman}

1363: Goodman, A. A., Benson, P. J., Fuller, G. A., Myers, P. C.,

1364: 1993, ApJ, 406, 528

1365: 

1366: \bibitem[\protect\citeauthoryear{Grady et al.}{2001}]{Grady}

1367: Grady, C. A., Polomski, E. F., Henning, Th., Stecklum, B., et al., 2001,

1368: AJ, 122, 3396 

1369: 

1370: \bibitem[\protect\citeauthoryear{Hartmann}{1998}]{Hartmann98b}

1371: Hartmann, L., 1998, Accretion Processes in Star Formation.

1372: Cambridge Univ. Press, Cambridge

1373: 

1374: \bibitem[\protect\citeauthoryear{Hartmann et al.}{2006}]{Hartmann}

1375: Hartmann, L., D'Alessio, P., Calvet, N., Muzerolle, J., 2006, ApJ, 648, 484

1376: 

1377: %\bibitem[\protect\citeauthoryear{Hartmann et al.}{1998}]{Hartmann98}

1378: %Hartmann, L., Calvet, N., Gullbring, E., D'Alessio, P., 1998, ApJ, 495, 385

1379: 

1380: %\bibitem[\protect\citeauthoryear{Klein et al.}{2003}]{Klein}

1381: %Klein, R., Apai, D., Pascucci, I., Henning, Th., Waters, L. B. F. M., 2003,

1382: %ApJ, 593, L57

1383: 

1384: \bibitem[\protect\citeauthoryear{Larson}{1984}]{Larson}

1385: Larson, R. B., 1984, MNRAS, 206, 197

1386: 

1387: \bibitem[\protect\citeauthoryear{Laughlin \& Bodenheimer}{1994}]{Laughlin}

1388: Laughlin, G., Bodenheimer, P., 1994, ApJ, 436, 335

1389: %

1390: %\bibitem[\protect\citeauthoryear{Lin \& Papaloizou}{1980}]{Lin}

1391: %Lin, D. N. C., Papaloizou, J., 1980, MNRAS, 191, 37

1392: 

1393: \bibitem[\protect\citeauthoryear{Lodato \& Rice}{2004}]{Lodato04}

1394: Lodato, G., Rice, W. K. M., 2004, MNRAS, 351, 630

1395: 

1396: \bibitem[\protect\citeauthoryear{Lodato \& Rice}{2005}]{Lodato05}

1397: \ul., 2005, MNRAS, 358, 1489

1398: 

1399: \bibitem[\protect\citeauthoryear{Lodato et al.}{2007}]{Lodato07}

1400: Lodato G., Meru F., Clarke C. J., Rice W. K. M., 2007, MNRAS, 374, 590

1401: 

1402: %\bibitem[\protect\citeauthoryear{Mohanty et al.}{2005}]{Mohanty}

1403: %Mohanty, S., Jayawardhana, R., Basri, G., 2005, ApJ, 626, 498

1404: 

1405: %\bibitem[\protect\citeauthoryear{Muzerolle et al.}{2003}]{Muzerolle}

1406: %Muzerolle, J., Hillenbrand, L., Calvet, N., Brice\~{n}o, C., Hartmann, L.,

1407: %2003, ApJ, 592, 266

1408: %

1409: %\bibitem[\protect\citeauthoryear{Muzerolle et al.}{2005}]{Muzerolle2}

1410: %Muzerolle, J., Luhman, K., Brice\~{n}o, C., Hartmann, L., Calvet, N.,

1411: %2005, ApJ, 625, 906

1412: 

1413: %\bibitem[\protect\citeauthoryear{Natta et al.}{2006}]{Natta}

1414: %Natta, A., Testi, L., Randich, S., 2006, A\&A, 452, 245 

1415: 

1416: %\bibitem[\protect\citeauthoryear{Padoan et al.}{2005}]{Padoan} 

1417: %Padoan, P., Kritsuk, A., Norman, M. L., Nordlund, A., 2005, ApJ, 622, L61

1418: 

1419: \bibitem[\protect\citeauthoryear{Pickett et al.}{2000}]{Pickett}

1420: Pickett, B. K., Cassen, P., Durisen, R. H., Link, R., 2000, ApJ, 529, 1034 

1421: 

1422: \bibitem[\protect\citeauthoryear{Polyachenko et al.}{1997}]{Polyachenko} 

1423:      Polyachenko V. L., Polyachenko E. V., Strel'nikov A. V., 1997, 

1424:      Astron.\ Zhurnal, 23, 598 (translated Astron.\ Lett.\ 23, 525) 

1425:      

1426: %\bibitem[\protect\citeauthoryear{Rice et al.}{2003}]{Rice03}

1427: %Rice, W. K. M., Armitage, P. J., Bate M. R., Bonnell I. A., 2003, MNRAS,

1428: %339, 1025

1429: 

1430: %\bibitem[1981]{Pringle}

1431: %Pringle, J. E., 1981, ARA\& A, 19, 137

1432: 

1433: \bibitem[\protect\citeauthoryear{Saigo \& Hanawa}{1998}]{Saigo}

1434: Saigo, K., Hanawa, T., 1998, ApJ, 493, 342

1435: 

1436: %\bibitem[\protect\citeauthoryear{Scholz \& Jayawardhana}{2006}]{Scholz2}

1437: %Scholz, A., Jayawardhana, R., 2006, ApJ, 638, 1056

1438: %

1439: \bibitem[\protect\citeauthoryear{Scholz et al.}{2006}]{Scholz1}

1440: Scholz, A., Jayawardhana, R., Wood, K., 2006, ApJ, 645, 1498

1441: 

1442: %\bibitem[\protect\citeauthoryear{Shu}{1977}]{Shu}

1443: %Shu, F. H., 1977, ApJ, 214, 488

1444: 

1445: \bibitem[\protect\citeauthoryear{Stone \& Norman}{1992}]{SN}

1446: Stone, J. M., Norman, M. L., 1992, ApJS, 80, 753

1447: 

1448: \bibitem[\protect\citeauthoryear{Tomley et al.}{1991}]{Tomley}

1449: Tomley, L., Cassen, P., Steiman-Cameron, T., 1991, ApJ, 382, 530

1450: 

1451: \bibitem[\protect\citeauthoryear{Toomre}{1981}]{Toomre}

1452:      Toomre, A., 1981, in Fall S. M., Lynden-Bell D., eds, The Structure

1453:      and Evolution of Normal Galaxies. Cambridge Univ. Press,

1454:      Cambridge

1455: 

1456: \bibitem[\protect\citeauthoryear{Turner et al.}{2007}]{Turner}

1457: Turner, N. J., Sano, T., Dziourkevitch, N., 2007, ApJ, 659, 729

1458: 

1459: \bibitem[\protect\citeauthoryear{Vorobyov \& Basu}{2005a}]{VB3}

1460: Vorobyov, E. I., Basu, S., 2005a, MNRAS, 360, 675

1461: 

1462: \bibitem[\protect\citeauthoryear{Vorobyov \& Basu}{2005b}]{VB1}

1463: \ul., 2005b, ApJ, 633, L137

1464: 

1465: \bibitem[\protect\citeauthoryear{Vorobyov \& Basu}{2006}]{VB2}

1466: \ul., 2006, ApJ, 650, 956

1467: 

1468: \bibitem[\protect\citeauthoryear{Vorobyov \& Theis}{2006}]{VT}

1469: Vorobyov, E. I., Theis, Ch., 2006, MNRAS, 373, 197

1470: 

1471: \bibitem[\protect\citeauthoryear{Weidenschilling}{1977}]{Weiden}

1472: Weidenschilling, S. J., 1977, ApSS, 51, 153

1473: 

1474: %\bibitem[\protect\citeauthoryear{Zhang}{1996}]{Zhang}

1475: %Zhang, X., 1996, ApJ, 457, 128

1476: 

1477: \end{thebibliography}

1478: 

1479: \appendix 

1480: \section{The Relative Role of Artificial Viscosity}

1481: 

1482: \subsection{Comparison with gravitational torques}

1483: 

1484: %A clear correlation between the rates of mass accretion and the strength of gravitational torques 

1485: %found in Fig.~\ref{fig5} and Fig.~\ref{fig6}

1486: %suggests that the gravitational torques drive the mass accretion.

1487: In principle, the torques due to artificial viscosity (present in our Eulerian numerical 

1488: hydrodynamics code) could also contribute to the radial transport of mass and angular momentum.

1489: To quantify this effect, we calculate 

1490: the net artificial viscosity torque in the disc ${\cal T}_{\rm av}$, defined as the sum of 

1491: local artificial viscosity torques 

1492: $\tau_{\rm av}(r,\phi)=-S(r,\phi)\, \partial P_{\rm av}/\partial \phi$ in the inner 600~AU.

1493: Here, $S(r,\phi)$ is the surface area occupied by a cell with polar coordinates $(r,\phi)$

1494: and $P_{\rm av}$ is the artificial viscosity pressure defined according to the usual 

1495: von Neumann \& Richtmyer prescription \citep[see e.g.][]{SN}.

1496: The net gravitational torque ${\cal T}_{\rm gr}$ is

1497: found by summing all local gravitational torques $\tau(r,\phi)$ in the inner 600~AU.

1498: We find that ${\cal T}_{\rm av}$ is at least ten orders of magnitude smaller than ${\cal T}_{\rm

1499: gr}$ and fluctuates near zero during the evolution, meaning that the positive and 

1500: negative artificial viscosity torques essentially cancel each other globally. 

1501: Furthermore, we must make sure that 

1502: %However, this tells us little about the magnitude 

1503: %of positive and negative artificial viscosity torques in the disc. If they are 

1504: %much larger than the corresponding gravitational torques, then 

1505: %any local disbalance in the negative and positive artificial viscosity

1506: even the negative part of the artificial viscosity torque cannot compete 

1507: with the corresponding negative gravitational torque and lead to 

1508: spurious radial mass transport.

1509: Hence, we also calculate the total {\it negative} torques due to gravity and artificial viscosity

1510: by summing only negative values of $\tau(r,\phi)$ and $\tau_{\rm av}(r,\phi)$ in the inner 600~AU, 

1511: respectively. Figure~\ref{fig8} shows these total negative torques at different evolutionary times.

1512: The negative torque due to gravity is clearly many orders of magnitude 

1513: greater than the negative torque due to artificial viscosity, especially in the

1514: late disc evolution when the spiral arms have disappeared. These tests enable us to conclude

1515: that artificial viscosity 

1516: has little influence on the radial transport of mass in our numerical simulations.

1517: 

1518: \begin{figure}

1519:   \resizebox{\hsize}{!}{\includegraphics{fig8.eps}}

1520:       \caption{Comparison of the total negative torques due to gravity (upper

1521:       curve) and artificial

1522:       viscosity (lower curve). Note that the total negative gravitational

1523:       torque is much greater than the total negative artificial viscosity

1524:       torque during the evolution.}

1525:          \label{fig8}

1526: \end{figure}

1527: 

1528: \subsection{Accretion without gravity}

1529: 

1530: A decisive test to confirm the principal role of gravitational torques is to 

1531: see if a non-self-gravitating disc can drive accretion rates comparable to those

1532: in the self-gravitating disc. To do this, we turn off self-gravity at $t=0.15$~Myr,

1533: i.e. just before disc formation but well after the central protostar has formed.

1534: Figure~\ref{fig9} shows the previously calculated $\mdot$ with self-gravity (black line) as well

1535: as the newly computed $\mdot$ in the absence of self-gravity (red line).

1536: The two rates are identical before the disc forms at $t\approx 0.17$~Myr.

1537: Afterwards, $\mdot$ in the non-self-gravitating

1538: disc declines quickly to a remarkably low value 

1539: $\dot{M}\approx 10^{-17}~M_\odot$~yr$^{-1}$ during 0.1~Myr, 

1540: confirming the principal role of self-gravity and gravitational 

1541: torques in driving the mass accretion in our model.

1542: We also note that it is not possible to follow the simulations 

1543: further in time, since the gas density near the inner boundary becomes negative. 

1544: This is because the accretion is then driven by small numerical imperfections of the inner 

1545: inflow computational boundary. As a result, the computational cells that are immediately

1546: adjacent to the inner boundary can become depleted of gas due to the absence of 

1547: gravitationally driven inward transport of gas from the outer disc. 

1548: 

1549: %Pressure force are negligible as compared to gravity forces (torques) ....

1550: 

1551: \begin{figure}

1552:   \resizebox{\hsize}{!}{\includegraphics{fig9.eps}}

1553:       \caption{Temporal evolution of the mass accretion rates $\dot{M}$

1554:       with (black line) and without (red line) self-gravity

1555:       of the disc.  Before  the disc formation at $t\approx 0.17$~Myr

1556:       both accretion rates are identical. After the disc formation, the

1557:       mass accretion rate in the non-self-gravitating disc declines quickly

1558:       to a negligibly small value.} 

1559: %This confirms the principal role of

1560: %self-gravity and gravitational torques in driving the mass accretion

1561: %      in our numerical simulations.

1562:          \label{fig9}

1563: \end{figure}

1564: 

1565: 

1566: 

1567: \end{document}

1568: 

1569: 

1570: 

1571: 

1572: 

1573: