1: %%

2: %% Beginning of file 'ms.tex'  Watanabe and Lin 2007 to ApJ

3: %%

4: %% Last Modified 09/07/2007

5: %%

6: %% This is a sample manuscript marked up using the

7: %% AASTeX v5.x LaTeX 2e macros.

8: 

9: 

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

11: %\documentclass[12pt,manuscript]{aastex}

12: % \documentclass{emulateapj}

13: 

14: %% definition of my own macros

15: 

16: \newcommand{\Msub}[1]{_{\rm #1}}       % roman subscript

17: \newcommand{\Msup}[1]{^{\rm #1}}       % roman superscript

18: \newcommand{\DP}[2]{\frac{\partial {#1}}{\partial {#2}}}   % partial differential

19: \newcommand{\Sc}[0]{{\cal S}{\rm c}} % Schumidt Number

20: \newcommand{\Punit}[1]{\, \mbox{#1}}   % Physical unit in the math mode

21: 

22: %% You can insert a short comment on the title page using the command below.

23: %\slugcomment{}

24: 

25: \shorttitle{Thermal Waves in Irradiated Protoplanetary Disks}

26: \shortauthors{Watanabe and Lin}

27: 

28: 

29: \begin{document}

30: 

31: \title{Thermal Waves in Irradiated Protoplanetary Disks}

32: 

33: \author{Sei-ichiro Watanabe\altaffilmark{1}}

34: \affil{Department of Earth and Planetary Sciences, Nagoya University,

35:     Chikusa, Nagoya 464-8601, Japan}

36: \email{seicoro@eps.nagoya-u.ac.jp}

37: \altaffiltext{1}{Visiting Researcher, UCO/Lick Observatory, 

38: University of California, Santa Cruz, CA 95064}

39: 

40: \and

41: 

42: \author{D. N. C. Lin\altaffilmark{2}}

43: \affil{UCO/Lick Observatory, University of California, Santa Cruz, CA 95064}

44: \email{lin@ucolick.org}

45: \altaffiltext{2}{KIAA, Peking University, Beijing 100871, China}

46: 

47: 

48: \begin{abstract}

49: Protoplanetary disks are mainly heated by radiation from the central

50: star. Since the incident stellar flux at any radius is sensitive to

51: the disk structure near that location, an unstable feedback may be

52: present. Previous investigations show that the disk would be stable to

53: finite-amplitude temperature perturbations if the vertical height of

54: optical surface is everywhere directly proportional to the gas scale

55: height and if the intercepted fraction of stellar radiation is

56: determined from the local grazing angle. We show that these

57: assumptions may not be generally applicable. Instead, we calculate

58: the quasi-static thermal evolution of irradiated disks by directly

59: integrating the global optical depths to determine the optical surface

60: and the total emitting area-filling factor of surface dust. We show

61: that, in disks with modest mass accretion rates, thermal waves are 

62: spontaneously and continually excited in the outer

63: disk, propagate inward through the planet-forming domains, and

64: dissipated at small radii where viscous dissipation is dominant. This

65: state is quasi-periodic over several thermal timescales and its

66: pattern does not depend on the details of the opacity law. The viscous

67: dissipation resulting from higher mass accretion stabilizes this 

68: instability such that an approximately steady state is realized throughout 

69: the disk. In passive protostellar disks, especially transitional disks, these waves

70: induce significant episodic changes in SEDs, on the time scales of

71: years to decades, because the midplane temperatures can vary by a

72: factor of two between the exposed and shadowed regions. The transitory

73: peaks and troughs in the potential vorticity distribution may also

74: lead to baroclinic instability and excite turbulence in the

75: planet-forming regions.

76: \end{abstract}

77: 

78: 

79: \keywords{accretion, accretion disks --- circumstellar matter --- instabilities 

80: --- planetary systems: protoplanetary disks --- stars: pre--main-sequence

81: --- solar system: formation}

82: 

83: 

84: \section{Introduction}

85: \label{SEC:Intro}

86: 

87: It has become widely accepted that dusty protoplanetary disks are

88: heated by radiation from the central star, and that this heating

89: mainly determines the physical structure of the outer regions of these

90: disks.  Observed infrared spectral energy distributions (SEDs) of the

91: T Tauri disks imply that their effective temperatures $T$ decreases

92: with disk radius $r$ more slowly than $T \propto r^{-3/4}$.  This

93: temperature distributions is usually explained by a model in which the

94: thermal structure of the disk is assumed to be geometrically flared,

95: i.e., the surface height $z\Msub{s}$ where stellar radiation is

96: absorbed curves away from the midplane (or equivalently $z\Msub{s}/r

97: \propto r^{\gamma}$, with $\gamma > 0$) \citep{ALS87}.

98: 

99: The outer regions of these disks are irradiated by the central star

100: \citep{KH87} and the flaring enables the disks to absorb more

101: radiation from the central star. In a steady state, the flaring index

102: $\gamma$ of a purely irradiated optically-thick disk can be obtained

103: from the balance between intercepted stellar flux $F\Msub{s}$ incident

104: on the surface at a low angle $\theta$ and emitted blackbody flux from

105: the disk interior, under the assumptions that 1) the surface height

106: $z\Msub{s}$ is everywhere proportional to the vertical gas scale

107: height $h$, 2) the intercepted fraction of stellar radiation is the

108: sine of the local grazing angle ($F\Msub{s} \propto \sin \theta$), and

109: 3) the central star is a point source. Under these circumstances,

110: there is a self-consistent power-law solution with $\gamma = 2/7$,

111: which corresponds to $T \propto r^{-3/7}$ \citep{KNH70, CG97} .

112: 

113: In addition to this particular power-law solution, there

114: exists an one-parameter family of solutions for the disk structure,

115: including the diverging aspect-ratio solutions and the asymptotically

116: conical (in which $z\Msub{s} \propto r$) solutions \citep{D00}.  Small

117: variations in the values of $z\Msub{s}$ at inner radii where the

118: integration starts can cause large differences in the disk structure

119: at large radii.  \citet{D00} speculated that such a sensitive nature

120: of the steady solutions is suggestive to an intrinsic instability

121: which must be analyzed with time-dependent governing equations.

122: 

123: More realistic steady solutions can be obtained numerically or

124: semi-analytically to take into account the effects of the finite

125: values of stellar radius, the disk optical depth, and the viscous

126: dissipation associated with the mass-accretion flow

127: \citep[e.g.,][]{CG97, CJC01, DDN01, THI05, DCHFS06, GL07}.  The most

128: important novel feature of these series of second-generation models is

129: the assumed presence of superheated surface dust layers above and

130: below the disk midplane \citep{CG97}.  Grains in these layers are

131: directly exposed to the stellar flux.  Grains much smaller than the

132: peak wavelength of the self emission are superheated because of their

133: low emissivity. The disk interior is heated by the superheated dust of

134: the layers rather than directly by the central star.

135: 

136: The two-layer disk model clearly explain the silicate and water-ice

137: emission bands in observed SEDs of Herbig Ae/Be stars and T Tauri

138: stars \citep{CJC01}.  But, in spite of their triumph in the modeling

139: of the observed SEDs, these models\footnote{Some author

140: \citep[e.g.,][]{DDN01, THI05, GL07} determined $\chi$

141: self-consistently, but these calculations are based on the

142: grazing-angle approximation and the assumption that $\chi$ changes

143: slowly with radius.} are based on the assumption that $z\Msub{s}$ is

144: proportional to the gas scale height $h$, with a fixed constant of

145: proportionality $\chi = z\Msub{s}/h = 4$ \citep{CJC01}.  While this

146: assumption has been justified by estimates which suggest that changes

147: of $\chi$ is small throughout the disk, the amount of dust in the

148: superheated layer is very sensitive function of $\chi$.  This

149: dependence arises because the dust spatial density at $z\Msub{s} =

150: \chi h$ is proportional to $\exp[-\chi^2/2]$.  Thus, the fixed-$\chi$

151: assumption may cause large discrepancy in determined values of

152: $z\Msub{s}$.

153:    

154: In previous analyses, the magnitude $F\Msub{s}$ is directly determined

155: from the grazing angle $\theta$ \citep{CJC01}.  This approximation

156: justified only in the case that the length of absorption layer (say,

157: where the optical depth to the starlight changes from 0.1 to 1) along

158: the starlight is smaller than the lengths of radial variations of

159: surface density or temperature.  This condition would not be satisfied

160: if the disk surface contains fluctuations resulting from the growth of

161: short-wavelength perturbations (see \S~\ref{SEC:Simple}).

162: 

163: In principle, $z\Msub{s}$ is determined by the condition that the

164: visual optical depth, which can be obtained by a direct integration

165: along the rays of starlight, is unity and the surface filling factor

166: $A\Msub{s}$ of the irradiated dust grains can be calculated through

167: the vertical (in the direction normal to the disk plane) integration

168: of geometrical opacity times mass density from $z\Msub{s}$ to

169: infinity. This global procedure yields $z\Msub{s}$ at any given radius

170: which depends not only on the local value of $h$ but also the disk

171: structure interior to that radius. Thus, $\chi = z\Msub{s}/h$ varies

172: both in space and time. Following this procedure, we can check for

173: self consistency by recalculating $z\Msub{s}$ based on the temperature

174: distribution obtained from the steady, constant-$\chi$ model. With

175: this inductive approach, we demonstrate that there are substantial

176: differences between the iterated values of $z\Msub{s}$ and the

177: initial, assumed values of $\chi h$. We also find that $A\Msub{s}$

178: calculated from the deduced values of $z\Msub{s}$ is substantially

179: different from the values of $\sin \theta$ extrapolated from the

180: constant-$\chi$ model.

181: 

182: Since the irradiation heating can play such a major role in

183: determining the vertical structure of protoplanetary disks, it is

184: important to investigate the stability of such disks against the

185: excitation of ripples on their surfaces. Under the assumption that the

186: thermal timescale is much longer than the dynamical timescale in

187: protostellar disks, \citet{DCH99} investigated the thermal stability

188: of the irradiation-dominated disks, using a simple cooling equation.

189: They found the vertically isothermal disk to be stable against finite

190: amplitude perturbations. The initial temperature perturbations

191: propagate inward and damp out quickly. However, in their analysis,

192: they assumed that $\chi$ is constant throughout the disk and

193: $F\Msub{s}$ is given by $\theta$ as in the grazing angle

194: approximation.  The inferred stability in that study may depend on

195: these assumptions. As we will see in \S~\ref{SEC:Simple}, changes in

196: $\chi$ may lead to an instability.

197: 

198: With a linear perturbation analysis, \citet{D00} showed that the

199: flaring disk solution may become unstable to infinitesimal

200: hydrodynamic perturbations when the cooling time of the disk is much

201: shorter than the dynamical time \citep[the opposite limit

202: of][]{DCH99}.  The amplitude of inwardly propagating waves grows

203: exponentially with a rate that is a decreasing function of the

204: wavelength.  Subsequently, \citet{DD04a} constructed a series of

205: numerical models to examine the two-dimensional structure and

206: evolution of protoplanetary disks around Herbig Ae/Be stars. In these

207: simulations, they studied the radiative transfer process under the

208: assumption that the disk always maintains a hydrostatic

209: equilibrium. (This assumption would not be appropriate for the limit

210: that the time scale of cooling is shorter than that of dynamics.)

211: They found two sets of asymptotically steady-state results which

212: include the monotonically flaring solutions and the self-shadowed

213: solutions. For the second set of solutions, the disk has a puffed-up

214: inner rim. They stated that their iteration procedure, in which the

215: hydrostatic equilibrium and the radiative transfer are treated

216: separately in alternate steps, may not, under some circumstances, lead

217: to a set of converged solutions. In their simulations, some wave-like

218: disturbances appear to propagate over the disk from one iterative step

219: to the next and these transitory features are never damped out

220: completely. Although they limited their presentations on disks around

221: Herbig Ae/Be stars, where the perturbation on the disk structure by

222: these waves are relatively minor, they revealed that this problem

223: appears to be more serious for disks around T Tauri stars.  We

224: speculate that this perturbation may be related to the above-mentioned

225: instabilities operating in the irradiation-dominated outer regions of

226: protostellar disks.

227: 

228: In this paper, we attempt to address two issues: 1) Are the steady-state 

229: solutions of irradiated disks constructed by the previous

230: one-dimensional models self-consistent and stable?  2) Do these regions

231: of disks tend to undergo quasi-periodic oscillations rather than

232: attain an asymptotic steady state? In principle, these questions

233: should be addressed with comprehensive two- or three-dimensional

234: numerical simulations. Such an approach is, however, fairly complicated,

235: time consuming, and often plagued with problems in the algorithm which

236: implements the radiative transfer processes \citep[e.g.,][]{DD04a}. 

237: Prior to these detailed simulations, it is useful to identify the dominant 

238: effects which regulate the dynamics of irradiated disks with a set of 

239: one-dimensional time-dependent analyses on the thermal evolution of 

240: protostellar disks.

241: 

242: Since thermal timescale is much shorter than viscous diffusion 

243: timescale, most previous studies adopted the assumption of thermal

244: equilibrium during the course of disks' global evolution.

245: Nevertheless, there have been a few investigations on the thermal

246: evolution of protoplanetary disks.  \citet{WNN90} investigated the

247: cooling and quasi-static contraction of the protoplanetary disks from

248: an initial high-temperature state.  They performed vertical

249: one-dimensional numerical calculations and found that the cooling

250: times are well estimated by a simple two-temperature (surface and

251: interior temperatures) prescription. In this paper, we utilize this

252: two-layer disk-temperature prescription to examine the stability and

253: thermal evolution of irradiated disks.

254: 

255: The simplest treatments for the thermal evolution of the irradiated

256: disk are radial one-dimensional models in which the

257: vertical structure of the disk at each radii is analyzed

258: independently.  In order to take into account of the irradiated

259: surface and the disk interior, we evaluate the surface height directly

260: from the location where the visual optical depth along the straight

261: lines from the star is unity. We show that such disks evolve to

262: quasi-periodic states in which thermal waves propagate inward through

263: intermediate disk radii, where planets are formed.

264: 

265: A simple discussion about the nature of thermal instability is given

266: in \S~\ref{SEC:Simple}. The basic assumption of our model and its

267: governing equations are presented in \S~\ref{SEC:Basic}. The results

268: of our numerical calculations are presented in \S~\ref{SEC:Results}

269: for both simple and realistic opacities.  Finally, we summarize our

270: findings and discuss some possible evolutionary scenarios.

271: 

272: \section{Simple discussion about thermal instability}

273: \label{SEC:Simple}

274: 

275: In this section, we discuss some potential causes for 

276: irradiation-dominated regions of disks to become thermally unstable.  

277: For illustration convenience, we adopt the following simplifying

278: assumptions: 1) the star is a point source, 2) the disk's internal

279: heat sources such as turbulent viscous heating as well as external

280: heating other than the stellar radiation are negligible, 3) the

281: optical depth of the disk is much larger than unity for both stellar

282: radiation and its own emission, and 4) the transport of energy in the

283: radial direction is much smaller than that in the vertical direction.

284: Note that these assumptions are adopted only in this section for the

285: purpose of pinpointing the physical process which leads to the thermal

286: instability.  All of these idealized assumptions will be relaxed in

287: the numerical simulations to be presented below.

288: 

289: Under these assumptions, we consider the heat balance in a 

290: geometrically thin disk which is irradiated by the central star.  

291: The energy equation reduces to

292: \begin{equation}

293: C \Sigma \DP{T\Msub{m}}{t} = 2 (F\Msub{s} - F\Msub{m}),

294: \label{EQ:therm}

295: \end{equation}

296: where $C$ is the specific heat per unit disk mass, $\Sigma$ is the

297: surface density of the disk, and $T\Msub{m}$ is the temperature of the

298: disk interior. Under assumption 2, the disk interior has an

299: approximately isothermal structure and $F\Msub{m}$ is the disk

300: blackbody emission given by

301: \begin{equation}

302: F\Msub{m} = \sigma T\Msub{m}^{4},

303: \end{equation}

304: and $F\Msub{s}$ is the intercepted stellar flux given by

305: \begin{equation}

306: F\Msub{s} = \frac{1}{2} \frac{L_\star}{4 \pi r^2} A\Msub{s},

307: \label{EQ:Fs1}

308: \end{equation}

309: where $\sigma$ is the Stefan-Boltzmann constant, $L_\star$ is the 

310: stellar luminosity, $r$ is the cylindrical

311: radial coordinate, and $A\Msub{s}$ is the total emitting area-filling

312: factor of superheated dust grains. The factor 1/2 in the right-hand

313: side of equation~(\ref{EQ:Fs1}) comes from the fact that surface

314: irradiated dust re-radiate half of the absorbed stellar flux toward

315: the disk interior (the rest toward infinity). Assuming a homogeneous

316: mixing of gas and dust, we can obtain $A\Msub{s}$ from (see

317: Appendix~\ref{SEC:Estimate})

318: \begin{equation}

319: A\Msub{s} = \tau\Msub{G} \mbox{ erfc} \left( \frac{z\Msub{s}}

320: {\sqrt{2} h} \right)

321: = \tau\Msub{G} \mbox{ erfc} \left( \frac{\chi}{\sqrt{2}} \right),

322: \label{EQ:Assec2}

323: \end{equation}

324: where $\mbox{erfc} (x)$ is the complimentary error function,

325: $\tau\Msub{G}$ is the geometrical optical depth of the disk midplane,

326: and $z\Msub{s}$ is the surface height where stellar radiation is

327: absorbed.  The ratio of the surface height $z\Msub{s}$ to the gas

328: scale height $h$ is denoted by $\chi \equiv z\Msub{s}/h$. In a

329: hydrostatic equilibrium, the gas scale height $h$ is given by

330: \begin{equation}

331: h = \frac{c\Msub{m}}{\Omega\Msub{K}} = \left( 

332: \frac{k\Msub{B} T\Msub{m} r^3}{\mu m\Msub{u} G M_\star}

333: \right)^{1/2},

334: \label{EQ:h}

335: \end{equation}

336: where $c\Msub{m}$ is the disk sound speed, $\Omega\Msub{K}$ is the Keplerian

337: angular velocity, $k\Msub{B}$ is the Boltzmann constant, $\mu$ is the

338: molecular weight of disk gas, $m\Msub{u}$ is the atomic mass unit, $G$

339: is the gravitational constant, and $M_\star$ is the stellar mass.

340: 

341: Figure~\ref{FIG:FmFs} displays the $F\Msub{m}$ and $F\Msub{s}$ (at $10

342: \Punit{AU}$) as functions of $T\Msub{m}$. Assuming that the initial

343: state is in a thermal equilibrium (corresponding to the point where

344: three thick lines cross) with $T\Msub{m} = T\Msub{m, eq}$, we impose a

345: small positive temperature perturbation.  We consider two extreme

346: cases: 1) If $z\Msub{s}$ is determined by the local disk structure,

347: $\chi = z\Msub{s}/h$ would retain a constant value during the increase

348: of $T\Msub{m}$ such that $A\Msub{s}$ would also be constant (see

349: eq.~[\ref{EQ:Assec2}]) and $F\Msub{s}$ would not change ({\it

350: dot-dashed line}). In this case the system would be stabilized because

351: $F\Msub{m} > F\Msub{s}$ for $T\Msub{m} > T\Msub{m, eq}$.  2) If

352: $z\Msub{s}$ is determined mostly by the attenuation by dust in the

353: inner regions of the disk, $z\Msub{s}$ would remain constant despite

354: changes in the local disk temperature and $h$ such that $A\Msub{s}$

355: would increase rapidly and $F\Msub{s}$ would increase much faster than

356: $F\Msub{m}$ ({\it thick dash curve}).  In this case the system would

357: be unstable because $F\Msub{m} < F\Msub{s}$ for $T\Msub{m} > T\Msub{m,

358: eq}$.

359: 

360: In their stability analysis, \citet{DCH99} assumed a constant $\chi$.

361: Based on the above analysis, this assumption naturally leads to stable

362: solutions.  In fact, most of the analysis on the structure of

363: irradiated disks are based on the constant-$\chi$ assumption

364: \citep[e.g.,][]{CJC01}.  The usual procedure to determine $z\Msub{s}$

365: is based on a geometrical consideration, i.e,

366: \begin{equation}

367: A\Msub{s} = \sin \theta \simeq \frac{z\Msub{s}}{r} 

368:                       \left( \frac{d \ln z\Msub{s}}{d \ln r} - 1 \right),

369: \label{EQ:Astheta}

370: \end{equation}

371: where $\theta$ is the grazing angle (i.e., the angle between the

372: starlight and the disk surface).  We refer this procedure to be the

373: grazing-angle approximation.  Most previous steady-state disk models

374: are constructed with equation~(\ref{EQ:Astheta}) under the assumption

375: that $\chi$ is constant throughout the disk.

376: 

377: However, the magnitude of $\chi$ is generally determined by the radial

378: structure of the disk as well as its local properties.  The surface

379: height $z\Msub{s}$ is determined by the optical depth integrated

380: through a ray of the stellar radiation.  We derive a ray integral and

381: check the validity of equation~(\ref{EQ:Astheta}) in

382: \S~\ref{SEC:Basic} and Appendix~\ref{SEC:Estimate}.  In principle,

383: equations~(\ref{EQ:Assec2}) and (\ref{EQ:Astheta}) must be resolved

384: simultaneously \citep{THI05}.  However, this set of equations is

385: fairly unstable to solve numerically because they do not contain

386: contributions which may reduce any steep temperature gradients in the

387: radial direction.  

388: 

389: Steep temperature gradient, if present, would invalidate the constant

390: $\chi$ and the grazing-angle approximations.  Physically, the radial

391: transport of heat suppresses the radial temperature gradient, but such

392: a process through the opaque regions of the disk must be analyzed with

393: multi-dimensional numerical simulations.  One of the most efficient

394: process of the radial heat transport is the radiative transfer from

395: the superheated dust grains at the surface of any radial location to

396: the disk midplane at adjacent radial regions.  Using a simple

397: one-dimensional model, we can take this oblique radiative transfer of

398: heat into account.

399: 

400: 

401: \section{Basic Equations}

402: \label{SEC:Basic}

403: 

404: Following the approaches of \citet{CG97} and \citet{GL07}, we

405: construct numerical models to study the thermal evolution of a

406: protostellar accretion disk.  The surface of the disk is illuminated

407: by the central star.  Exposed to the stellar radiation, sub-mm dust

408: grains in the surface layers of the disk are superheated.  We consider

409: the case that dust mass of the disk is so large that the disk midplane

410: (except for an innermost region where silicates are evaporated) is

411: optically thick to the stellar radiation.  In contrast to the previous

412: section, the heat sources for the disk interior in these numerical

413: models include both irradiation from the superheated grains on the disk

414: surface and the viscous dissipation associated with the accretion

415: flow.  We adopt a cylindrical coordinate system $(r, \phi, z)$ in

416: which the $z=0$ plane represents the disk midplane and the origin is

417: at the location of the central star.  Since the star-and-disk system

418: is symmetric with respect to the midplane, we describe our results for

419: the upper half of the disk only.

420: 

421: In order to simplify the problem we adopt the two-layer axisymmetric

422: disk model proposed by \citet{CG97}.  In this model, the disk consists

423: of a surface superheated layer where the dust temperature is

424: $T\Msub{s} (r)$ and a disk interior where the dust and gas temperature

425: is assumed to be uniform at $T\Msub{m} (r)$.  This model is simple to

426: use and includes all the essential ingredient to analyze the onset,

427: evolution, and stabilization of thermal instability in protostellar

428: disks.  However, such a simplification would be invalid if the disk

429: optical depth $\tau\Msub{m}(T\Msub{m})$ to its intrinsic radiation is

430: much larger than unity and viscous heating rate is larger than surface

431: heating rate.  However, the dust optical depth may be self-limited by

432: their rapid growth through cohesive collisions, so that

433: $\tau\Msub{m}(T\Msub{m}) \la 10$ throughout the disk \citep[see Fig.~6

434: in][]{THI05}.  Thus, the two-layer model is valid even in the inner

435: disk where dust surface density is larger.

436: 

437: The two-layer model is invalid at the inner edge of the disk, where

438: disk is irradiated not only from the top but from the radial

439: direction.  The disk may have a puffed-up inner rim

440: \citep[e.g.,][]{DD04a}, but a set of two-dimensional radiative

441: transfer calculations is needed to determine the structure of the

442: innermost region. In this work we simply assume that disk within $0.1

443: \Punit{AU}$ is optically thick in the radial direction and has no

444: puffed-up rim that might cast shadows over the outer regions of the

445: disk.  We confine our calculations only in the regions $r > 0.1

446: \Punit{AU}$, where the two-layer model is valid.

447: 

448: The thermal timescale of the disk interior at radius $r$ is given by

449: (see eq.~[\ref{EQ:energy}])

450: \begin{eqnarray}

451: t\Msub{th} & = & \frac{(\gamma\Msub{a}+1)}{2(\gamma\Msub{a}-1)} 

452:              \frac{c\Msub{m}^2 \Sigma}{\sigma T\Msub{m}^4}

453:              \nonumber \\

454:           & \simeq & 53 \left(\frac{\Sigma_0}{\Sigma\Msub{H0}} \right)

455:                      \left( \frac{T\Msub{m0}}{124 \Punit{K}} \right)^{-3}

456:                      \left( \frac{r}{1 \Punit{AU}} \right)^{3q-p} \Punit{yr},

457: \label{EQ:tth}

458: \end{eqnarray}

459: where $\gamma\Msub{a}$ is the adiabatic exponents and $c\Msub{m}$ is

460: the sound speed of the disk interior.  For evaluation, we assume here

461: power-law distributions for the total ($\mbox{gas} + \mbox{dust}$) surface 

462: density $\Sigma(r) \propto r^{-p}$ and the midplane temperature 

463: $T\Msub{m}(r) \propto r^{-q}$.  The normalization factors, $\Sigma_0$ and

464: $T\Msub{m0}$, refer to their corresponding values at $r =1\Punit{AU}$.  

465: The nominal value of surface density is given by that in the minimum mass 

466: solar-nebula (MMSN) model in which $\Sigma\Msub{H} = 1.7 \times

467: 10^3 \Punit{g} \Punit{cm}^{-2}$ \citep{H81}.

468: 

469: We assume that the thermal timescale $t\Msub{th}$ to be much longer

470: than the dynamical time ($\Omega\Msub{K}^{-1}$), but much shorter 

471: than the viscous evolution time ($r^2/\nu$, where $\nu$ is the turbulent

472: viscosity).  In this case we can regard that the whole region of the

473: disk is always in a hydrostatic equilibrium in the vertical direction

474: and has time-independent surface densities.

475: 

476: The temperature $T\Msub{s}$ of superheated dust grains is given by

477: \begin{equation}

478:  \frac{L_\star}{4 \pi r^2} = 4 \epsilon\Msub{s} \sigma T\Msub{s}^4,

479: \label{EQ:Ts}

480: \end{equation}

481: where $\epsilon\Msub{s}$ is the averaged emissivity of the dust grains

482: at $T\Msub{s}$.  Along a ray from the surface of the star, the

483: superheated layer extends outward until the position where visual

484: optical depth has reached unity.  We take into account the attenuation 

485: of the stellar photons by defining the height $z\Msub{s}$ of the bottom 

486: of the superheated layer with the following equation

487: \begin{equation}

488: \tau\Msub{s}(T_\star; r, z\Msub{s}(r)) = 1.

489: \label{EQ:zsdef}

490: \end{equation}

491: Here $\tau\Msub{s} (T_\star; r, z)$ is the optical depth between the

492: central star and the point $(r, z)$ to the blackbody radiation peaked

493: at the stellar effective temperature $T_\star$, given by the following

494: integration

495: \begin{equation}

496:  \tau\Msub{s}(T_\star; r, z) = \int_{R_\star}^r \bar{\kappa}\Msub{s}(T_\star)

497:  \rho\Msub{d}(r',\zeta r') \left( 1 + \zeta^2 \right)^{1/2} \, dr'

498: \label{EQ:taus}

499: \end{equation}

500: along a straight path from the star to the point $(r,z)$.  Here

501: $R_\star$ is the stellar radius, $\zeta \equiv z/r$ is the aspect ratio,

502: $\rho\Msub{d}(r',z')$ is the spatial mass density of dust at $(r', z')$,

503: and $\bar{\kappa}\Msub{s}(T\Msub{rad})$ is the Planck mean opacity of

504: the grains interacting with the blackbody radiation peaked at

505: $T\Msub{rad}$.  Note that we define here the grain opacity per unit

506: {\it dust mass}, not per unit total ($\mbox{gas} + \mbox{dust}$) mass as in the usual

507: definition, because it is convenient for the consideration of the case

508: that the dust-to-gas ratio may change vertically.  The emissivity in

509: equation~(\ref{EQ:Ts}) can be given by

510: \begin{equation}

511:  \epsilon\Msub{s} = \frac{\bar{\kappa}\Msub{s}(T\Msub{s})}{%

512:                                                        \bar{\kappa}\Msub{s}(T_\star)}.

513: \label{EQ:epsilons}

514: \end{equation}

515: 

516: The energy equation includes heating from both stellar irradiation and

517: viscous dissipation as well as radiative losses from the disk surface

518: such that \citep[see, e.g.,][]{WNN90}

519: \begin{equation}

520:  \frac{(\gamma\Msub{a} + 1)}{2(\gamma\Msub{a}-1)} 

521:  \frac{k\Msub{B}\Sigma}{\mu m\Msub{u}} 

522:  \DP{T\Msub{m}}{t} = 

523:     2 \left[ F\Msub{s} - F \Msub{m} \right] 

524:       + \frac{3}{4 \pi} \dot{M} \Omega\Msub{K}^2,

525: \label{EQ:energy}

526: \end{equation}

527: where $\dot{M}$ is the mass accretion rate, which we assume to be

528: constant throughout the disk. Note that the steady state assumption is

529: compatible with a power-law surface density distribution for some

530: effective viscosity prescriptions \citep{CG97, DCHFS06, GL07}.

531: 

532: Further, $F\Msub{s}$ and $F\Msub{m}$ are, respectively, the thermal

533: radiation fluxes downward from the superheated dust grains high up in

534: the disk atmosphere and upward from dust grains in the disk interior,

535: \begin{equation}

536:  F\Msub{s}(r) = \left[ 1 - e^{-2\tau\Msub{m}(T\Msub{s})} \right] 

537:                   \frac{L_\star}{8 \pi} 

538:                   \left\langle \frac{A\Msub{s}}{r^2} + 

539:                     \frac{4R_\star}{3 \pi r^3} \right\rangle, 

540: \label{EQ:Fs}

541: \end{equation}

542: %

543: \begin{equation}

544:  F\Msub{m}(r) = \left[ 1 - e^{-2\tau\Msub{m}(T\Msub{m})} \right] \sigma 

545:  T\Msub{m}^4, 

546: \label{EQ:Fm}

547: \end{equation}

548: where $\tau\Msub{m}(T\Msub{s})$ and $\tau\Msub{m}(T\Msub{m})$ are the

549: optical depths of the disk interior (from $z=0$ to $z=z\Msub{s}$) to

550: the radiation from the superheated dust grains and to its own

551: emission, respectively, and $A\Msub{s}$ is the total emitting-area

552: filling-factor of superheated dust grains.  We consider the effect of

553: finite radius of the central star in $F\Msub{s}$, which is important

554: in the inner part of the disk.  We also consider the effects of

555: oblique radiative transfer: the angular brackets in the right-hand

556: side of equation~(\ref{EQ:Fs}) represent radial averaging of radiation

557: emitting from superheated dust within the adjacent regions (see

558: \S~\ref{SEC:Results}).  The factors 2 in the exponential functions in

559: equations~(\ref{EQ:Fs}) and (\ref{EQ:Fm}) also denote oblique

560: radiative transfer in the disk interior \citep{THI05}.

561: 

562: Once the dust density distribution $\rho\Msub{d}(r,z)$ of the disk is

563: specified, $A\Msub{s}$ and $\tau\Msub{m}$ can be determined from the

564: following integration:

565: \begin{equation}

566:  A\Msub{s}(r) = 1 - \exp \left[ - \int_{z\Msub{s}(r)}^{\infty} 

567: \tilde{\kappa}\Msub{s} (T_\star)

568:                                          \rho\Msub{d}(r,z') \, dz' \right],

569: \label{EQ:As}

570: \end{equation}

571: %

572: \begin{equation}

573:  \tau\Msub{m}(T\Msub{rad}; r) = \int_0^{z\Msub{s}(r)} 

574:       \bar{\kappa}\Msub{m}(T\Msub{rad})

575:                                          \rho\Msub{d}(r,z')  \, dz',

576: \end{equation}

577: where $\bar{\kappa}\Msub{m}(T\Msub{rad})$ is the Planck mean

578: opacity of midplane grains interacting the blackbody radiation peaked

579: at temperature $T\Msub{rad}$.  Further details of dust opacity are

580: shown in \S \ref{SEC:Results} and Appendix \ref{SEC:Opacity}.

581: 

582: We assume that the total ($\mbox{gas} + \mbox{dust}$) surface density 

583: $\Sigma$ of the disk is a simple power-law distribution in the radial 

584: direction:

585: \begin{equation}

586:  \Sigma = \Sigma_0 \left( \frac{r}{r_0} \right)^{-p},

587: \label{EQ:Sigma}

588: \end{equation}

589: which is kept constant during the thermal evolution considered in this

590: work.

591: 

592: Taking the effects of dust settling, we can obtain dust density 

593: distribution $\rho\Msub{d}$ of the disk in the two-temperature model 

594: adopted here (see Appendix~\ref{SEC:DustDensity}). However, 

595: computationally intense iterative calculation is needed to determine, 

596: self-consistently, the magnitudes of $\rho\Msub{d}$ and $z\Msub{s}$ 

597: simultaneously.  Instead of equations~(\ref{EQ:rhodm}) and 

598: (\ref{EQ:rhods}), we adopt, in most of the calculations, the following 

599: simple density distribution for the dust:

600: \begin{equation}

601:  \rho\Msub{d} (r,z) = \frac{\Sigma\Msub{d}}{\sqrt{2 \pi} h}  

602: \exp\left( -\frac{z^2}{2 h^2} \right),

603: \label{EQ:rhodsimple}

604: \end{equation} 

605: where $\Sigma\Msub{d}$ is the surface density of dust. We put

606: $\Sigma\Msub{d} = f\Msub{d} \Sigma$, where $f\Msub{d}$ is the

607: dust fraction in the surface density.  We dub  $f\Msub{d}$ the dust-to-gas 

608: ratio.  In the disk interior, dust sedimentation is not so important unless

609: the radii of dust are not so large that we can formally set the dust

610: density distribution to be equation~(\ref{EQ:rhodsimple}).  In the 

611: surface layer, dust settling reduces the dust density while high 

612: surface temperature $T\Msub{s}$ raise it, so that 

613: equation~(\ref{EQ:rhodsimple}) also gives a good estimate. In some cases 

614: we compared results with more realistic dust distribution given in 

615: Appendix~\ref{SEC:DustDensity} and found that they are very similar

616: unless dust sizes are not so large.  We discuss the difference of the 

617: results between the two distributions in \S~\ref{SEC:Discuss}.

618: 

619: 

620: We found that if  the gradient $d \ln z\Msub{s} / d \ln r$

621: changes rapidly in $r$ direction, the approximation used to derive 

622: equation~(\ref{EQ:Astheta}) is no longer justified (see 

623: Appendix~\ref{SEC:Estimate}).  For this reason we use 

624: equations~(\ref{EQ:zsdef}) and (\ref{EQ:As})

625: instead of equation~(\ref{EQ:Astheta}).

626: 

627: \section{Numerical Results}

628: \label{SEC:Results}

629: 

630: We adopt the following values as fixed parameters for all models: the

631: mass, radius, and luminosity of the central star are set to be

632: $M_\star = 1 M_\sun$, $R_\star = 2.085 R_\sun$, and $L_\star =

633: 1 L_\sun$, respectively, so that its effective temperature is $T_\star

634: = 4000 \Punit{K}$.  For the disk gas, we specify $\mu = 2.34$ and

635: $\gamma\Msub{a}=1.4$.

636: 

637: We adopted the phenomenological MMSN model \citep{H81} for the 

638: standard gas and dust surface density

639: distribution. In this model, $\Sigma_0 = \Sigma\Msub{H0} = 1.7 \times

640: 10^3 \Punit{g} \Punit{cm}^{-2}$ with $r_0 = 1 \Punit{AU}$ and $p =

641: 1.5$ in equation~(\ref{EQ:Sigma}).  For comparison,

642: we also calculate a relatively flat $\Sigma$ distribution with $p =

643: 1.0$ and $\Sigma_0 = 3.54 \times 10^2 \Punit{g} \Punit{cm}^{-2}$.

644: These surface densities are kept constant with time. 

645: The dust-to-gas ratio $f\Msub{d}$

646: and solid material density $\rho\Msub{mat}$ are set to be 0.014 and

647: $1.4 \Punit{g} \Punit{cm}^{-3}$, respectively.  In \S~\ref{SS:CO} we

648: neglect any changes of dust surface density due to sublimation

649: and use the value of $f\Msub{d}$ throughout the disk.  In \S~\ref{SS:RO}

650: we consider the effect of ice sublimation. We also vary the

651: mass accretion rate from $\dot{M} = 0$ to $ 10^{-7} M_\sun

652: \Punit{yr}^{-1}$.  These steady-state accretion rates are consistent

653: with our specified surface density and temperature distributions

654: provided the magnitude of $\alpha$ is a function of the radius,

655: where $\alpha$ is the non-dimensional turbulent viscosity in the

656: so-called $\alpha$-prescription \citep{SS73}.

657: 

658: The normalization time unit is the thermal timescale given by

659: equation~(\ref{EQ:tth}) with $\Sigma = \Sigma_0$ and $T\Msub{m} =

660: T\Msub{m0}$.  We denote the time unit as $t\Msub{th,0}$ and the

661: non-dimensional time as $\hat{t}=t/t\Msub{th,0}$.  For the standard model, we set the

662: magnitude of $\Sigma_0 = \Sigma\Msub{H0}$ and $T_0 = 124 \Punit{K}$,

663: so that $t\Msub{th,0} = 53 \Punit{yr}$.  Note that the local thermal

664: timescale $t\Msub{th}$ is nearly constant with $r$ in the standard

665: disk model with $p=1.5$.  The non-dimensional time step used in the numerical

666: integration is set to be $\delta \hat{t} = 0.005$.  In order to verify

667: numerical convergence, we also perform several calculations with time

668: step half of the standard value and confirmed that the results have no

669: significant changes.

670: 

671: The spatial grid consists of 90 points (the standard case) or 180

672: points (the high-resolution case), logarithmically distributed,

673: between $r = 0.1 \Punit{AU}$ and $100 \Punit{AU}$.  Numerical

674: oscillation would be induced if there is no radial exchange of energy

675: at all.  At any radius, the disk interior is exposed not only to the

676: superheated surface grains directly overhead, but also obliquely to

677: those at adjacent radial locations.  Such a radial exchange of energy

678: tends to suppresses instabilities for short-wavelength perturbations.

679: In our numerical scheme, we assume that the isotropic radiation comes

680: from all radius $r'$ within $|r'-r| < z\Msub{s}(r)$ contributes to the

681: heating at radius $r$ as expressed in equation~(\ref{EQ:Fs}).  This

682: implementation stabilizes the short wave-length oscillation and the

683: results are essentially independent of the numerical resolution.

684: 

685: \subsection{Constant Opacity}

686: \label{SS:CO}

687: 

688: In this subsection, we first illustrate the dominant features using a

689: simple opacity model.  We adopt the emissivity and opacity (per unit

690: {\it dust mass}) of the grains interacting with blackbody radiation peaked

691: at temperature $T_i$ to be

692: \begin{equation}

693:  \epsilon\Msub{s}(T_i)  = \left( \frac{T_i}{T_\star} \right)^\beta \mbox{ and }

694:  \bar{\kappa}\Msub{s}(T_i) = \bar{\kappa}\Msub{s0} 

695:                                        \left( \frac{T_i}{T_\star} \right)^\beta,

696: \end{equation} 

697: where we choose the value of $\bar{\kappa}\Msub{s0} = 10^2

698: \Punit{cm}^{2} \Punit{g}^{-1}$, which approximately corresponds to the

699: commonly defined opacity per unit {\it gas mass} with the value of $1

700: \Punit{cm}^{2} \Punit{g}^{-1}$.  Most of the calculation shown here is

701: for $\beta = 0$ but we also calculate some models with $\beta = 1$ for

702: comparison purposes. 

703: 

704: For initial conditions, we adopt the steady state solution obtained

705: from the time integration with fixing $\chi(r)$.  This set of initial

706: conditions does not correspond to the asymptotic steady-state

707: solutions because the initial estimate of $\chi$ is not

708: self-consistently compatible with the actual aspect ratio

709: $\zeta\Msub{s}$ of the surface.

710: Nevertheless, the numerical calculations relax to nearby steady

711: solutions if they exist. In order to verify that our results are

712: independent of the adopted initial conditions, we calculate the

713: evolution of the disk with several different initial guesses for

714: $\chi$ and found that the system reaches to the same asymptotic state.

715: 

716: We first show the results of the calculations with no mass accretion

717: ($\dot{M} = 0$).  Figure~\ref{FIG:init-ck-M0-T} shows the initial

718: evolution of the midplane temperature $T\Msub{m}$ as well as the

719: surface temperature $T\Msub{s}$, which is kept constant with time. The

720: elapsed time is $\hat{t} = 1.6$, i.e., $t = 1.6 t\Msub{th,0} \simeq 85

721: \Punit{yr}$ and each curve corresponds to a time step $\Delta \hat{t}

722: = 0.2$ ($\Delta t = 10.6 \Punit{yr}$).  At first, the initial state is

723: almost stable in the innermost region and the outermost region.  But,

724: in the intermediate regions (0.5--$20 \Punit{AU}$), the disk becomes

725: unstable.  At a typical instance of time, four local temperature peaks

726: (high $T\Msub{m}$ and $z\Msub{s}$) coexist and are amplified. These

727: peaks move inward (toward the star) as they grow.  Hence, we refer

728: these propagating perturbations as waves.  During the amplification of

729: the waves, they cast shadows over the outer regions of the peak. The

730: temperature $T\Msub{m}$ decreases in the shadowed regions.  Each

731: fully-grown wave has a sharp slope on the inner `exposed side' of the

732: peak and a gentler decline on the outer `shadow side'.  The waves

733: propagate inward with velocities about a few tenth of $r/t\Msub{th}$.

734: The outermost wave ($\sim 16 \Punit{AU}$) begins to grow just outside

735: the shadowed region of the inner adjacent wave when the shadowed

736: region is developed. This tendency shows that the outermost wave may

737: be induced by the wave ahead of it.

738: 

739: Additional time integration shows that as these waves propagate inward

740: they begin to decay when they reach inside $1 \Punit{AU}$.  The waves

741: are completely damped out at around $0.25 \Punit{AU}$.  In contrast, new

742: waves are formed continually at the outermost region ($> 20

743: \Punit{AU}$) of the computational domain.  The growth region of waves

744: gradually retreats and the maximum amplitude of each wave during its

745: propagation cycle gradually increases with time.  The system reaches a

746: quasi-periodic state when $\hat{t} \sim 8$.  Figure~\ref{FIG:ck-M0-T}

747: shows the evolution of $T\Msub{m}$ at this stage.  Waves are

748: continuously formed and amplified at the outer disk ($>

749: 30~\Punit{AU}$), then propagate inward with nearly constant amplitude

750: through the intermediate disk regions, begin to decay at about 1~AU,

751: and are damped out completely at around $0.25 \Punit{AU}$.  Temperature

752: in the innermost disk region ($r < 0.25 \Punit{AU}$) attains steady

753: values. The propagation speed of the waves is approximately given by

754: $r/t\Msub{th}$.  In the intermediate disk radii (1--$20 \Punit{AU}$),

755: the peak temperature of each wave is 2--3 times higher than the bottom

756: temperature in the inner adjacent shadowed region.

757: 

758: The radial profile of each wave is somewhat skewed.  The half width of

759: individual waves is about 0.1--$0.2r$ on the inner side and

760: 0.2--$0.4r$ on the outer side.  The wavelength is approximately twice

761: as large as $z\Msub{s}(r)$. Due to the steep radial temperature

762: gradient, the magnitude of $z\Msub{s}(r)$ is affected by the variation

763: in the thickness $h$ at the disk regions interior to $r$ on this

764: length scale.  The ratio of the radii between two adjacent wave peaks

765: are about 20 if both waves are within $r > 1 \Punit{AU}$.  Near the

766: inner boundary of the propagating-wave zone, the finite size of the

767: star ($R_\star$) becomes comparable to the surface height $z\Msub{s}$.

768: This time-independent contribution in equation~(\ref{EQ:Fs})

769: essentially stabilizes the innermost region of the disk.

770: 

771: The change of other variables at the same epoch as 

772: Figure~\ref{FIG:ck-M0-T} are shown in 

773: Figures~\ref{FIG:ck-M0-zeta}--\ref{FIG:ck-M0-Pv}.  

774: Figure~\ref{FIG:ck-M0-zeta} shows the time

775: evolution of $\zeta\Msub{s} = z\Msub{s}/r$.  Outside $0.25

776: \Punit{AU}$, $\zeta\Msub{s}$ changes stepwise.  The two or three steep

777: jumps of $\zeta\Msub{s}$ correspond to the leading inner side of the

778: thermal waves. The magnitude of increase in $z\Msub{s}$ at the leading

779: edge of each step reaches a maximum of 1.5 around $10 \Punit{AU}$.

780: The flat portions of $\zeta\Msub{s}$ correspond to the shadowed

781: regions, where $\zeta\Msub{s}$ is determined by the stellar ray which

782: passes above the peak of each wave.

783:    

784: Figure~\ref{FIG:ck-M0-zovh} shows the time evolution of $\chi =

785: z\Msub{s}/h = \zeta\Msub{s}/\zeta\Msub{h}$.  The local minima of

786: $\chi$ are located just ahead of the peaks in $T\Msub{m}$.  The

787: decrease of $\chi$ is essential for the temperature raise.  In the

788: shadowed regions $\chi$ increases because $\zeta\Msub{s}$ is kept

789: almost constant (see Fig.~\ref{FIG:ck-M0-zeta}), whereas $\zeta\Msub{h}

790: = h/r$ decreases.  Changes in the value of $\chi$ produces the

791: variations in the surface filling factor $A\Msub{s}$.  The time

792: evolution of $A\Msub{s}$ is shown Fig.~\ref{FIG:ck-M0-As}.  Sharp

793: peaks of $A\Msub{s}$, which correspond to the minima of $\chi$ (see

794: eq.~[\ref{EQ:Aschi}]), propagate inward.  Note that the variation in

795: the amplitude of $A\Msub{s}$ is very large, ranging from a few tenth

796: at the peaks to below $10^{-3}$ in the shadowed region.  Such large

797: changes of $A\Msub{s}$ induce rapid heating in front of the waves and

798: rapid cooling in the shadowed regions. The unperturbed $z\Msub{s}$ in

799: the intermediate regions is comparable to the half width of the

800: propagating waves.  Modest variations in $\chi$ can lead to nonlinear

801: dissipation, such that the wave amplitudes are also limited in these

802: region.

803: 

804: We also plot the evolution of the distribution of the logarithmic

805: pressure gradient $d {\rm ln} P/ d{\rm ln} r$

806: (Fig.~\ref{FIG:ck-M0-dP}), where $P(r)$ is the pressure in the

807: midplane of the disk.  This plot indicates that the gas pressure

808: gradient is nearly reversed just in front of the peak of the waves.

809: This inversion occurs due to the steep positive temperature gradient

810: in the exposed, leading, inner face of the waves.  Consequently, the

811: velocity of the gas departs significantly from its unperturbed

812: sub-Keplerian values.  In a follow-up paper, we will consider the

813: associated gas drag on grains of various sizes. Another interesting

814: quantity is the distribution of the potential vorticity (or

815: vortensity):

816: \begin{equation}

817: \frac{\Omega\Msub{ep}}{\Sigma} = \frac{1}{\Sigma r^3}

818: \frac{d}{dr} r^4 \Omega^2,

819: \end{equation}

820: where $\Omega\Msub{ep}$ is the epicyclic frequency and $\Omega$ is the

821: angular velocity of gas.  Note that in a quasi-Keplerian disk

822: $\Omega\Msub{ep} \simeq \Omega\Msub{K} \propto r^{-1.5}$, so that the

823: potential vorticity is almost constant for MMSN with $p=1.5$. The

824: thermal waves disturb the potential vorticity through the change of

825: pressure gradient.  Figure~\ref{FIG:ck-M0-Pv} displays the evolution

826: of the distribution of the potential vorticity.  There are peaks and

827: troughs around the waves.  Local extrema of this quantity can lead

828: baroclinic instabilities which may excite turbulence

829: \citep[e.g.][]{KB03} in the dead zone where the magneto-rotational

830: instability may have limited influences \citep{G96}.  Further

831: investigation of this possibility will also be considered elsewhere.

832: 

833: Next, we examine the dependence of wave excitation and propagation on

834: the mass accretion rate ($\dot{M}$).  Other than the value of

835: $\dot{M}$ in the energy equation, we adopt the same model parameters

836: as for the no-accretion case shown in

837: Figures~\ref{FIG:init-ck-M0-T}--\ref{FIG:ck-M0-Pv}.  Thus, in the

838: present context, the primary physical effects associated with the

839: accretion flow is the viscous dissipation. This internal energy source

840: (viscous dissipation) in equation~(\ref{EQ:energy}) has a greater

841: fractional contribution to the energy budget in the inner regions than

842: the outer regions of the disk \citep{GL07}.

843: 

844: Although quasi-periodic oscillations are excited in all cases, the

845: radial extent where they propagate to depends on the magnitude of

846: $\dot{M}$.  Figure~\ref{FIG:ck-M8-T} shows the time evolution of

847: $T\Msub{m}$ in a quasi-periodic state for the case of $\dot{M} =

848: 10^{-8} M_\sun \Punit{yr}^{-1}$.  The result is similar to that

849: obtained by neglecting the viscous dissipation. In this case, waves

850: are excited in the outer region, propagate inward and are damped out

851: in the innermost region.  In comparison with the no-accretion model,

852: the innermost region in this case is hotter ($T\Msub{m}(0.1

853: \Punit{AU}) \sim 600 \Punit{K}$) with larger thickness.  The {\it

854: time-independent} contribution of viscous heating to the energy 

855: equation provides a stabilizing effect to a slightly larger radial 

856: extent ($r < 0.5 \Punit{AU}$) compared with the no-accretion 

857: case ($T\Msub{m}(0.1 \Punit{AU}) \sim 500 \Punit{K}$ and 

858: $r < 0.25 \Punit{AU}$).

859: 

860: The quasi-periodic state, however, is drastically changed in the

861: $\dot{M} = 1 \times 10^{-7} M_\sun \Punit{yr}^{-1}$ model

862: (Fig.~\ref{FIG:ck-M7-T}).  The disk becomes stable interior to about

863: $6 \Punit{AU}$.  In the outer region, there are two high-temperature

864: peaks, both oscillate quasi-periodically. The positions of the two

865: peaks do not coherently propagate inward as in the case of $\dot{M} =

866: 1 \times 10^{-8} M_\sun \Punit{yr}^{-1}$, but they fluctuate to and

867: fro around 12~AU and 40~AU, respectively.  The ranges of temporal

868: temperature changes are less than 2~K over the disk, so that the

869: disk is regarded to be in an approximately steady state.  This result

870: shows that the disk becomes stabilized as $\dot{M}$ increases.

871: 

872: Finally, we show the dependence on the disk surface density $\Sigma$.

873: Note that $\Sigma$ affects the evolution not only through the optical

874: depth of the surface layer but through the thermal timescale.  We

875: perform several calculations for $\Sigma_0 = 3.54 \times 10^2

876: \Punit{g} \Punit{cm}^{-2}$ and $p = 1.0$.  Note that the time unit for

877: this case is $t\Msub{th,0} = 11.0 \Punit{yr}$.  We calculate the disk

878: evolution for this surface density distribution with several values of

879: $\dot{M}$ and find that the results are quite similar to the case with

880: the standard surface density distribution.  The quasi-periodic wave

881: solutions are obtained for small $\dot{M}$, whereas the disk becomes

882: nearly steady for large $\dot{M}$.

883: 

884: Figure~\ref{FIG:ck-M8-p10-T} shows the temporal variation of

885: $T\Msub{m}$ in the quasi-periodic state for this prescribed

886: $\Sigma(r)$ distribution with $\dot{M} = 10^{-8} M_\sun

887: \Punit{yr}^{-1}$.  Comparing to the case with $p=1.5$

888: (Fig.~\ref{FIG:ck-M8-T}), the wave propagating region is relatively

889: narrow, confined to the regions between 1~AU and 20~AU, and the

890: maximum amplitude of waves is also smaller.  Outermost region where $r

891: > 20 \Punit{AU}$, the amplitudes of waves are limited and do not

892: exceed about 2~K.  The amplitudes grow as waves propagate from 20~AU

893: to 8~AU, then they attain a nearly constant value from 8~AU to 1.5~AU.

894: The waves are finally damped at 1.5--1~AU. In an analogous inviscid

895: model (i.e., with an identically prescribed $\Sigma (r)$

896: distribution but without viscous dissipation), the result is quite

897: similar to that in Figure~\ref{FIG:ck-M8-p10-T} except that the wave

898: propagating region extends slightly closer to the star.  The results

899: for the $\dot{M} = 10^{-7} M_\sun \Punit{yr}^{-1}$ model show that the

900: disk attains an approximately steady state with only very small fluctuations

901: (amplitude is less than a few tenth~K) remaining in the outermost

902: region.

903: 

904: All the above results are for $\beta = 0$. We also calculate cases

905: with $\beta =1$.  Other than a modification in the distribution of

906: $T\Msub{s}$ and $T\Msub{m}$, the time dependent nature of wave

907: excitation and propagation is essentially independent of the value of

908: $\beta$.  In all cases, we find that the quasi-periodic nature of the

909: inwardly-propagating thermal waves is realized for the disk with

910: $\dot{M} \la 10^{-8} M_\sun \Punit{yr}^{-1}$.  The basic features of

911: this state do not strongly depend on other parameters such as $p$ and

912: $\beta$.  In order to verify the universality of these results, we

913: perform further calculations with more realistic opacities.

914: 

915: \subsection{Realistic Opacities}

916: \label{SS:RO}

917: 

918: In this subsection, we present the results based on models with more realistic 

919: grain-opacity prescription given by \citet{THI05}.  In this

920: prescription, grains are assumed to be consisted of a uniform mixture

921: of H$_2$O-ice, organics, olivine, pyroxene, metallic iron, and

922: troilite, the abundances of which are given by \citet{PHB94}.  For

923: relatively high temperatures ($ T > 160 \Punit{K}$), we use the dust

924: opacity of grains without ice or organics.  For the low-temperature

925: ($T < 160 \Punit{K}$) state, the opacity includes the contribution

926: from ice and organics grains.  Based on the single-sized monochromatic

927: opacity table made by H.~Tanaka \citep{THI05}, we calculate the Planck

928: mean of the size-averaged opacities $\bar{\kappa}\Msub{s}(T\Msub{rad})$ 

929: and $\bar{\kappa}\Msub{m}(T\Msub{rad})$ with an assumed power-law

930: grain-size distribution $n(s) \propto s^{-3.5}$, which is truncated

931: for a minimum and maximum range of $s\Msub{min} = 0.1 \Punit{\micron}$

932: and $s\Msub{max} = 1 \Punit{mm}$, respectively (see Appendix \ref{SEC:Opacity}).

933:  

934: For illustrative models, we adopt the standard surface-density

935: distribution of the MMSN model.  For the

936: gas-to-dust ratio $f\Msub{d}$, we consider two limiting cases: 1) a

937: constant ratio $f\Msub{d} = 0.014$ and 2) a ratio which includes the

938: effect of ice sublimation.  In the latter case, $f\Msub{d} = 0.0043$

939: for $T > 160 \Punit{K}$ and $f\Msub{d} = 0.014$ for $T < 160

940: \Punit{K}$.  In the actual implementation, these values are smoothly

941: connected around the sublimation temperature by a hyperbolic tangent

942: function.  Note that even in the constant-$f\Msub{d}$ cases, the

943: opacity undergoes a transition across $T=160 \Punit{K}$.

944: 

945: We perform a set of calculations with this realistic opacity.  Similar

946: to the previous models with constant opacity, the disk evolves into a

947: quasi-periodic state after about 10 times of the thermal timescale.

948: First, we show the results for the case with a constant gas-to-dust

949: ratio.  Figure~\ref{FIG:vk-M8-T} shows the time evolution of

950: $T\Msub{m}$ in the asymptotic quasi-periodic state of the disk with

951: $\dot{M} = 1 \times 10^{-8} M_\sun \Punit{yr}^{-1}$.  The surface

952: temperature $T\Msub{s}$ is also shown in Figure~\ref{FIG:vk-M8-T}.

953: From equations~(\ref{EQ:Ts}) and (\ref{EQ:epsilons}), we note that

954: $T\Msub{s}$ is independent of time.  Comparing with the 

955: constant-opacity case in which $\beta = 0$ (Fig.~\ref{FIG:ck-M8-T}),

956: $T\Msub{s}$ is everywhere larger for the realistic grain

957: opacity, because of the effect of superheating. There is a small jump 

958: at the radius where opacity law changes

959: ($T\Msub{s} \simeq 160 \Punit{K}$).  For the same value of

960: $\dot{M}$, the structure of wave propagation region is very similar

961: to that of the constant-opacity case.

962: 

963: With a sufficiently large mass-accretion rate ($\dot{M} = 1 \times

964: 10^{-7} M_\sun \Punit{yr}^{-1}$), the disk with a realistic opacity is

965: also stabilized by the effect of viscous dissipation (Fig.~\ref{FIG:vk-M7-T}).

966: There are three local maxima in the $T\Msub{m}$ distribution.  The

967: innermost peak at about $6 \Punit{AU}$ corresponds to an opacity

968: transition in the disk's surface layer ($T\Msub{s} \sim 160

969: \Punit{K}$).  All these local peaks do not propagate but fluctuate

970: quasi-periodically with a small amplitude.

971: 

972: Finally, we show the effect of ice sublimation.

973: Figure~\ref{FIG:vk-ev-M8-T} displays the evolution of $T\Msub{m}$ for

974: the model in which the dust surface density is modified by the ice

975: sublimation.  Taking into account the effect of dust's size sorting on

976: the opacity, we assume that surface dust grains has a smaller maximum

977: size $s\Msub{s,min}$ than those in disk interior $s\Msub{m,min}

978: =s\Msub{min} =1 \Punit{mm}$.  We choose 

979: $s\Msub{s,min} = 1 \Punit{\micron}$.

980: Owing to the increase of $A\Msub{s}$ associated with the ice

981: condensation in surface layer, the $T\Msub{m}$ distribution attains a

982: local maximum near $5 \Punit{AU}$.  This peak does not propagate in

983: time.  Interior to this snow line, a wave propagates to about $0.5

984: \Punit{AU}$.  Outside this snow line, there is another local maximum

985: near $20 \Punit{AU}$, which may be induced by the emergence of the

986: first peak.  Compare to the model in which the ice sublimation is

987: neglected (Fig.~\ref{FIG:vk-M8-T}), the ice-condensation induces the

988: formation of a local thermal maximum which prevents waves from

989: emerging at large radii and propagate inward.

990: 

991: \section{Summary and Discussion}

992: \label{SEC:Discuss}

993: 

994: We have performed a set of radial one-dimensional calculations to

995: examine the thermal evolution of hydrostatic disks, using the direct

996: integration of optical depths $\tau\Msub{s}(T_\star)$ to determine the

997: optical surface $z\Msub{s}$ and total emitting area-filling factor

998: $A\Msub{s}$ of a superheated layer.  Our results suggest that, in

999: regions with modest and steep radial temperature gradients, the

1000: constant $\chi = z\Msub{s}/h$ assumption is incompatible with the

1001: computed height of the surface where $\tau\Msub{s}(T_\star; r,

1002: z\Msub{s}) =1$.  The initial state obtained by a fixed-$\chi$

1003: iteration evolves spontaneously to the state where thermal waves

1004: grows.  The disks evolve to a quasi-periodic state where thermal waves

1005: continuously propagate toward the star in the intermediate radii.

1006: 

1007: The driving mechanism for this thermal instability is the intense

1008: stellar irradiation high up in the disk atmosphere.  It is a

1009: consequence of a ``shadowing effect'' in which the surface where most

1010: of the stellar photons are intercepted at any given radial location

1011: may be affected by the vertical structure in the disk regions interior

1012: to that radius.  This quasi-periodic state is stabilized by viscous

1013: dissipation associated with the mass accretion flow through the disk.

1014: For the cases of $\dot{M} = 10^{-7} M_\sun \Punit{yr}^{-1}$ wave

1015: excitation and propagation are suppressed and the disks reach

1016: approximately steady states. In order to eliminate the possibility of

1017: artificial dependence on the initial conditions, we perform the

1018: following numerical experiments. By setting the initial condition for

1019: an approximately steady state with $\dot{M} = 10^{-7} M_\sun

1020: \Punit{yr}^{-1}$ and by decreasing the mass accretion rate to $\dot{M}

1021: = 10^{-8} M_\sun \Punit{yr}^{-1}$, we calculate the time evolution of

1022: the disk.  We find that the system evolves to a quasi-periodic state

1023: within about 10 times of thermal timescale. The asymptotic

1024: quasi-periodic state is identical to that obtained with the standard

1025: calculation for $\dot{M} = 10^{-8} M_\sun \Punit{yr}^{-1}$, which is

1026: shown in Figure~\ref{FIG:ck-M8-T}.  Inversely, we also start a

1027: calculation from a quasi-periodic state with $\dot{M} = 10^{-8} M_\sun

1028: \Punit{yr}^{-1}$ and increase the mass accretion rate to $\dot{M} =

1029: 10^{-7} M_\sun \Punit{yr}^{-1}$.  The disk evolves into an

1030: approximately steady state, which is almost identical to that obtained

1031: with the standard calculation for $\dot{M} = 10^{-7} M_\sun

1032: \Punit{yr}^{-1}$, which is shown in Figure~\ref{FIG:ck-M7-T}.  These

1033: results show that the state of the disks are determined by the

1034: instantaneous mass accretion rate (note that viscous evolution time is

1035: much longer than thermal timescale).

1036: 

1037: The result that the quasi-periodic wave-propagating states exist in

1038: disks with modest disk accretion rate is robust to variations in the

1039: disks' structural parameters such as their surface density profile,

1040: opacity law, and vertical dust distribution.  Whereas these parameters

1041: weakly affect the positions of the inner and outer boundaries of the

1042: wave-propagating domains, they do not affect the basic features of

1043: the waves such as their amplitude and shape.

1044: 

1045: We also perform some calculations for the constant-opacity case using

1046: a general surface density profile (eq.~[\ref{EQ:rhods}] instead of

1047: eq.~[\ref{EQ:rhodsimple}]).  The results are quite similar with that

1048: of the simple density profile.  The difference is even less noticeable

1049: than that due to change of opacity law.

1050: 

1051: Using the calculated thermal structure of the disk, we obtain the disk

1052: SED.  We assume that the disk is face-on and

1053: truncated both at inner $0.1 \Punit{AU}$ and outer $100 \Punit{AU}$.

1054: Figure~\ref{FIG:vk-ev-M8-SED} shows the evolution of the SEDs

1055: associated with the model illustrated in Figure~\ref{FIG:vk-ev-M8-T}.

1056: Contribution from the central star to the SEDs is included in this

1057: figure.  The SEDs are expected to oscillate periodically within

1058: mid-infrared wavelengths.  This variation correspond to the periodic

1059: propagation of the thermal waves.  In this case, the period of

1060: oscillation is about $ t\Msub{th,0} \simeq 53 \Punit{yr}$. In general

1061: this time scale varies depending on the surface density distribution

1062: and accretion rate in the disk as well as the luminosity of the

1063: central star. The change of disk SED comes from not only emission from

1064: the disk interior where $T\Msub{m}$ changes but also emission from the

1065: superheated surface layer where, even if $T\Msub{s}$ is kept constant,

1066: $A\Msub{s}$ would modulate.  The surface emission also contributes to

1067: the water-ice and silicate emission bands.  We expect the relative

1068: heights of these emission bands in SEDs to modulate with time.  Note

1069: that this calculated SED is somewhat artificial because disk is

1070: truncated at the inner and outer edges. Thus, the direct comparison

1071: with observation may be meaningless.  Nevertheless, the predicted

1072: relative variations of SEDs due to propagation of the thermal waves

1073: are likely to be important in the interpretation of the observed SEDs,

1074: especially those of the so-called transitional disks taken by {\it

1075: SPITZER} Infrared Spectrograph (IRS) \citep[e.g.,][]{FHC06}. 

1076:    

1077: In this work, we assume $t\Msub{th}$ to be much longer than

1078: $\Omega\Msub{K}^{-1}$, so that a hydrostatic equilibrium would be

1079: quickly re-established after the passage of the thermal waves.  This

1080: assumption is invalid in the outer regions (say $> 20 \Punit{AU}$,

1081: see eq.~[\ref{EQ:tth}]).  But our calculations show that the wave 

1082: propagating region is extended to regions interior to $1 \Punit{AU}$.  

1083: This result implies that even if dynamic effect may suppress the 

1084: thermal waves in the outer disk, the waves can still be excited in 

1085: the inner regions. According to previous linear analysis

1086: \citep{D00}, the disk becomes unstable to infinitesimal hydrodynamic

1087: perturbations when $t\Msub{th} \ll \Omega\Msub{K}^{-1}$.  This type 

1088: of instabilities may also excite the thermal waves.

1089: 

1090: The results presented here is based on simple radial one-dimensional

1091: analysis in which a two-temperature approximation is adopted to

1092: describe the vertical structure of the disk.  We also neglected

1093: modulations in the surface density and accretion rate throughout the

1094: disk.  Our next task is to relax these assumptions and to generalize

1095: our results to a set of genuine two-dimensional simulations in which

1096: the radiation and mass transfer can be considered simultaneously.

1097: Since dust growth time or radial migration time are comparable to the

1098: thermal timescale, it will also be important to consider the evolution

1099: of the dust particles \citep[e.g.,][]{TL03, DD04b}.  These

1100: investigations will be carried out in the future and presented

1101: elsewhere.

1102: 

1103: \acknowledgments

1104: 

1105: We wish to thank the anonymous referee for valuable comments.

1106: We thank to Dr. H.~Tanaka for providing us with the single-sized

1107: monochromatic opacity table used in \citet{THI05}.  We are grateful to

1108: Drs P.~Garaud and K.~Kretke for useful conversation.  This work is

1109: supported in part by NASA (NAGS5-11779, NNG04G-191G, NNG06-GH45G), 

1110: JPL(1270927), NSF(AST-0507424, PHY99-0794), and Grant-in-Aid of the

1111: Japanese Ministry of Education, Science, and Culture (19540239).

1112: 

1113: \appendix

1114: 

1115: \section{Opacities}

1116: \label{SEC:Opacity}

1117: 

1118: The Planck mean opacity is given by

1119: \begin{equation}

1120:  \bar{{\kappa}}_j(T\Msub{rad}) = \frac{\int_0^\infty \tilde{\kappa}_\nu (T_j) 

1121:              B_\nu(T\Msub{rad}) \, d\nu}{\int_0^\infty B_\nu (T\Msub{rad}) \, d\nu}, 

1122: \end{equation}

1123: where subscript $j$ represents ``s'' (surface) or ``m'' (disk interior),

1124: $\nu$ is the frequency, $B_\nu (T\Msub{rad})$ is the Planck function, and 

1125: \begin{equation}

1126:  \tilde{\kappa}_\nu (T_j) = \frac{\int_{s\Msub{min}}^{s\Msub{max}} 

1127:              \kappa_\nu (T_j, s) s^3 n(s) \, ds}{\int_{s\Msub{min}}^{s\Msub{max}}

1128:              s^3 n(s) \, ds}.

1129: \end{equation}

1130: Here $n(s) \, ds$ is the number density of grains with radii between

1131: $s$ and $s+ds$, and the size distribution has the lower cutoff

1132: $s\Msub{min}$ and the upper cutoff $s\Msub{max}$.  In addition,

1133: $\kappa_\nu (T, s)$ is the single-sized ($s$) monochromatic ($\nu$)

1134: dust opacity of grains with temperature $T$.

1135: 

1136: We assume a power-law size distribution $n(s) \propto s^{-3.5}$, which

1137: is truncated for a minimum and maximum range of $s\Msub{min} = 0.1

1138: \Punit{\micron}$ and $s\Msub{max} = 1 \Punit{mm}$.

1139: 

1140: The grain compositions and optical constants are adopted from

1141: \citet{THI05} and references therein.  We use the resultant opacity

1142: table given by H. Tanaka. The table gives two single-sized

1143: monochromatic dust opacities as functions of $\nu$ and $s$: one for

1144: grains without ice or organics at high temperatures of $ T > 160

1145: \Punit{K}$ and the other for grains including ice or organics at $T <

1146: 160 \Punit{K}$.

1147: 

1148: \section{Exact Solution of Dust Density Distribution}

1149: \label{SEC:DustDensity}

1150: 

1151: The gas density distribution $\rho\Msub{g}$ of the two-temperature 

1152: model is given by

1153: \begin{equation}

1154:  \rho\Msub{g} (r,z) =  \left\{ 

1155:             \begin{array}{ll}

1156:                 \rho\Msub{g}(r,0) \exp\left( -\frac{z^2}{2 h^2} \right)&

1157:                 \mbox{if $|z| \le z\Msub{s}$} \\

1158:                 \rho\Msub{g}(r,z\Msub{s}) \exp\left( -\frac{z^2 - 

1159:                   z\Msub{s}^2}{2 H^2} \right)

1160:                 & \mbox{if $|z| \ge z\Msub{s}$,}

1161:             \end{array}

1162:                              \right.

1163: \label{EQ:rhog}

1164: \end{equation}

1165: where $h = c\Msub{m} \Omega\Msub{K}^{-1}$ (see eq.~[\ref{EQ:h}]) and

1166: $H = c\Msub{s} \Omega\Msub{K}^{-1}$ ($c\Msub{s}$ is the sound speed in

1167: the surface layer) are the gas scale height in disk interior and in

1168: the surface layer, respectively.

1169: 

1170: In a steady state, the sedimentation flux of the dust grains with

1171: their terminal velocity balances the diffusive mass flux due to gas

1172: turbulence \citep[see eq.~{[30]} in][]{TL02} such that:

1173: \begin{equation}

1174: -\rho\Msub{d} \Omega\Msub{K} \hat{t}\Msub{stop} z = 

1175:   \frac{\rho\Msub{g} \nu}{\Sc} 

1176:   \frac{\partial \left(\rho\Msub{d}/\rho\Msub{g} \right)}{\partial z}, 

1177: \label{EQ:diffusive}

1178: \end{equation}

1179: where $\hat{t}\Msub{stop}$ is the stopping time normalized by 

1180: $\Omega\Msub{K}^{-1}$, $\nu$ 

1181: is the turbulent viscosity, and $\Sc$ is the Schmidt number 

1182: representing coupling strength between grains and gas. The nondimensional

1183: stopping time $\hat{t}\Msub{stop}$ is given by

1184: \begin{equation}

1185: \hat{t}\Msub{stop} = \frac{\rho\Msub{mat} s \Omega\Msub{K}}{\rho\Msub{g}c\Msub{t}},

1186: \label{EQ:tstop}

1187: \end{equation}

1188: where $\rho\Msub{mat}$ is the material mass density, $s$ is the dust

1189: radius, and $c\Msub{t}$ is the mean thermal velocity. 

1190: 

1191: Solving equation~(\ref{EQ:diffusive}) with equations~(\ref{EQ:rhog})

1192: (in the case of $|z| \le z\Msub{s}$) and (\ref{EQ:tstop}), we obtain the

1193: dust density distribution of the disk interior to be \citep[see

1194: eq.~{[31]} in][]{TL02}

1195: \begin{equation}

1196:  \rho\Msub{dm} (r,z) =  \rho\Msub{dm}(r,0) \exp\left[ -\frac{z^2}{2 h^2} 

1197:                             - \frac{\Sc \hat{t}\Msub{stop,m}}{\alpha} 

1198:                               \left( \exp \frac{z^2}{2h^2} - 1 \right) 

1199:                           \right],

1200: \label{EQ:rhodm}

1201: \end{equation}

1202: where $\hat{t}\Msub{stop,m}$ is the stopping time in the midplane

1203: and $\alpha = \nu/(c\Msub{m} h)$ is the non-dimensional 

1204: turbulent viscosity in the so-called $\alpha$-prescription \citep{SS73}.  The 

1205: dimensionless stopping time in the midplane is given by

1206: \begin{equation}

1207:  \hat{t}\Msub{stop,m} = \frac{\pi}{2} \frac{\rho\Msub{mat} s}{\Sigma}.

1208: \end{equation}

1209: The midplane dust density $\rho\Msub{dm}(r,0)$ is determined

1210: by the vertical integration of equation~(\ref{EQ:rhodm}) to be the

1211: surface density of dust $\Sigma\Msub{d} = f\Msub{d} \Sigma$, where

1212: $f\Msub{d}$ is the dust fraction in surface density. 

1213: 

1214: Solving equation~(\ref{EQ:diffusive}) with equations~(\ref{EQ:rhog}) 

1215: (in the case of $|z| \ge z\Msub{s}$) and (\ref{EQ:tstop}), we obtain 

1216: the dust density distribution in the surface layer

1217: \begin{equation}

1218:  \rho\Msub{ds} (r,z)  = 

1219:     \rho\Msub{dm}(r,z\Msub{s}) 

1220:     \exp\left[ -\frac{z^2- z\Msub{s}^2}{2 H^2} 

1221:                   - \frac{\Sc \hat{t}\Msub{stop,m} h}{\alpha H}

1222:                      \exp \left( \frac{z\Msub{s}^2}{2 h^2} \right) 

1223:                      \left(  \exp \frac{z^2-z\Msub{s}^2}{2H^2} - 1 \right)

1224:            \right],

1225: \label{EQ:rhods}

1226: \end{equation}

1227: where we put $\rho\Msub{ds}(r,z\Msub{s}) = \rho\Msub{dm}(r,z\Msub{s})$

1228: and assume that $\Sc$ and $\alpha$ in the surface layer have the same values

1229: as those in the disk interior.

1230: 

1231: In the disk interior, dust sedimentation is not so important unless

1232: the radii of dust are not so large that we can formally set

1233: $\rho\Msub{dm} = \rho\Msub{d}$.  In the surface layer, dust settling

1234: reduces the dust density while high surface temperature $T\Msub{s}$

1235: raise it, so that $\rho\Msub{d}$ also gives a good estimate.  Thus,

1236: instead of equations~(\ref{EQ:rhodm}) and (\ref{EQ:rhods}), we adopt, 

1237: in most of the calculations, the simple density distribution given by 

1238: equation~(\ref{EQ:rhodsimple}).

1239: 

1240: \section{The validity of the grazing-angle approximation}

1241: \label{SEC:Estimate}

1242: 

1243: We derive the grazing-angle

1244: approximation (eq.~[\ref{EQ:Astheta}]) and discuss its validity.

1245: Substituting $\rho\Msub{d} (r,z)$ in equation~(\ref{EQ:As}) with that

1246: in equation~(\ref{EQ:rhodsimple}) and assuming $A\Msub{s} \ll 1$, we

1247: obtain

1248: \begin{equation}

1249: A\Msub{s}

1250:   = \tau\Msub{v} \mbox{ erfc} \left( \frac{z\Msub{s}}{\sqrt{2} h} \right)

1251:   = \tau\Msub{v} \mbox{ erfc} \left( \frac{\chi}{\sqrt{2}} \right),

1252: \label{EQ:Aschi}

1253: \end{equation}

1254: where $\tau\Msub{v} = \tilde{\kappa}\Msub{s}(T_\star) \Sigma\Msub{d}/2$ and 

1255: $\chi = z\Msub{s}/h$.  

1256: 

1257: We derive the relation of $A\Msub{s}$ with a path integration from 

1258: the star to the point $(r,z)$.  Using equations~(\ref{EQ:taus}) and 

1259: (\ref{EQ:rhodsimple}), equation~(\ref{EQ:zsdef}) can be written as

1260: \begin{equation}

1261: \tau\Msub{s}(T_\star; r, z\Msub{s}(r)) 

1262:  = \tau\Msub{v} \left( 1+\zeta\Msub{s}^2 \right)^{1/2}

1263: \sqrt{\frac{2}{\pi}} \int_{R_\star}^r \frac{\hat{\Sigma}\Msub{d}'}{h(r')}

1264: e^{-\chi'^2/2} \, dr' = 1

1265: \end{equation}

1266: with $\zeta\Msub{s} = z\Msub{s}/r$, $\hat{\Sigma}\Msub{d}' = 

1267: \Sigma\Msub{d}(r')/\Sigma\Msub{d}(r)$, and $\chi' = \zeta\Msub{s} r'/h(r')$.

1268: Here we assume that $\bar{\kappa}\Msub{s}(T_\star)$ is independent of

1269: position of the path. 

1270: Substituting the integral variable from $r'$ to $\chi'$ and noting that 

1271: $\chi' \gg 1$ for $r' = R_\star$, we obtain 

1272: \begin{equation}

1273: \tau\Msub{v} \left( 1+\zeta\Msub{s}^2 \right)^{1/2} \zeta\Msub{s}^{-1}

1274: \sqrt{\frac{2}{\pi}} \int_{\chi}^\infty \hat{\Sigma}\Msub{d}'

1275: \left( \frac{d \ln \zeta'_h}{d \ln r'} \right)^{-1} e^{-\chi'^2/2} \, 

1276: d\chi' = 1,

1277: \label{EQ:taucond}

1278: \end{equation}

1279: where we define $\zeta_h=h/r$ and $\zeta'_h = h(r')/r'$.

1280: Here we assume $\chi'$ is a monotonically increasing function with $r'$.

1281: From equations~(\ref{EQ:Aschi}) and (\ref{EQ:taucond}), we obtain

1282: \begin{equation}

1283: A\Msub{s} =  

1284: \left( 1 + \zeta\Msub{s}^2\right)^{-1/2} \zeta\Msub{s}

1285: \left\langle \hat{\Sigma}\Msub{d}' \left( 

1286: \frac{d \ln \zeta'_h}{d \ln r'} 

1287: \right)^{-1} \right\rangle_{g(\chi'),\chi}^{-1}

1288: \end{equation}

1289: with

1290: \begin{equation}

1291: g(\chi') = \left[ \mbox{erfc } \left( \chi/\sqrt{2} \right) \right]^{-1}

1292:              \sqrt{2/\pi} \exp \left(-\chi'^2/2 \right),

1293: \end{equation}

1294: where $\langle X \rangle_{f(x'),x} = \int_x^\infty X(x') f(x')\, dx'$

1295: represents the weighted average with a weight function $f(x')$.  

1296: 

1297: If we substitute the Gaussian weight $g(\chi')$ with the delta function 

1298: $\delta(\chi'-\chi)$ and assume $\zeta\Msub{s} \ll 1$, we obtain

1299: \begin{equation}

1300: A\Msub{s} =  \zeta\Msub{s}

1301: \left( \frac{d \ln \zeta_h}{d \ln r} \right).

1302: \label{EQ:Asaprox}

1303: \end{equation}

1304: Except for a small difference, this equation corresponds to 

1305: equation~(\ref{EQ:Astheta}).  

1306: However, if  the gradient $d \ln \zeta_h / d \ln r$

1307: changes rapidly in $r$ direction, the approximation used to derive 

1308: equation~(\ref{EQ:Asaprox}) is no longer justified. 

1309: In this case $A\Msub{s}$ is determined not only by the local gradient

1310: of $\zeta_h$ but by the gradients in inner regions 

1311: because the weight function $g(\chi')$ is extended to the inner radii.

1312: For this reason we use equations~(\ref{EQ:zsdef}) and (\ref{EQ:As})

1313: instead of equation~(\ref{EQ:Asaprox}).

1314: 

1315: \begin{thebibliography}{}

1316: %

1317: \bibitem[Adams, Lada, \& Shu(1987)]{ALS87} Adams, F. C., Lada, C. J., \&

1318:     Shu, F. H. 1988, \apj, 312, 788 

1319: \bibitem[Chiang \& Goldreich(1997)]{CG97} Chiang, E. I., \&

1320:     Goldreich, P.  1997, \apj, 490, 368

1321: \bibitem[Chiang et al.(2001)]{CJC01} Chiang, E. I., Joung, M. K.,

1322:     Creech-Eakman, M. J., Qi, C., Kessler, J. E., Blake, G. A.,

1323:     \& van Dishoeck, E. F. 2001, \apj, 547, 1077

1324: \bibitem[D'Alessio et al.(1999)]{DCH99} D'Alessio, P.,

1325:     Cant\'{o}, J., Hartmann, L., Calvet N., \& Lizano, S.  1999, \apj,

1326:     511, 896

1327: \bibitem[D'Alessio et al.(2006)]{DCHFS06} D'Alessio, P., Calvet N., 

1328:     Hartmann, L., Franco-Hern\'{a}ndez \& Serv\'{i}n, H. 2006, \apj,

1329:     638, 314

1330: \bibitem[Dullemond(2000)]{D00} Dullemond, C. P. 2000, \aap,

1331:     361, L17

1332: \bibitem[Dullemond \& Dominik(2004a)]{DD04a} Dullemond, C. P. \& 

1333:     Dominik, C. 2004a, \aap, 417, 159 

1334: \bibitem[Dullemond \& Dominik(2004b)]{DD04b} Dullemond, C. P. \&

1335:     Dominik, C. 2004b, \aap, 421, 1075

1336: \bibitem[Dullemond et al.(2001)]{DDN01} Dullemond, C. P., Dominik, C.,

1337: \& Natta, A. 2001, \apj, 560, 957

1338: \bibitem[Furlan et al.(2006)]{FHC06} Furlan, E., et al.

1339: 2006 \apjs, 165, 568

1340: \bibitem[Gammie(1996)]{G96} Gammie, C. F. 1996, \apj, 457, 355    

1341: \bibitem[Garaud \& Lin(2007)]{GL07} Garaud, P., \&

1342:     Lin, D. N. C.  2007, \apj, 654, 606

1343: \bibitem[Hayashi(1981)]{H81} Hayashi, C. 1981, Prog.~Theor.~Phys.~Suppl.,

1344:     70, 35

1345: \bibitem[Kenyon \& Hartmann(1987)]{KH87} Kenyon, S. J., \& Hartmann, L.

1346:     1987, \apj, 323, 714

1347: \bibitem[Klahr \& Bodenheimer(2003)]{KB03} Klahr, H. H. \& Bodenheimer,

1348: P. 2003, \apj, 582, 869

1349: \bibitem[Kusaka et al.(1970)]{KNH70} Kusaka, T., Nakano, T., \& 

1350: Hayashi, C. 1970, Prog.~Theor.~Phys.~Suppl., 44, 1580 

1351: \bibitem[Pollack et al.(1994)]{PHB94} Pollack, J. B., Hollenbach, D.,

1352:     Beckwith, S., Simonelli, D. P., Roush, T., \&

1353:     Fong, W.  1994, \apj, 421, 615

1354: \bibitem[Shakura \& Sunyaev(1973)]{SS73} Shakura, N. I., \&

1355:     Sunyaev, R. A.  1973, \aap, 24, 337

1356: \bibitem[Takeuchi \& Lin(2002)]{TL02} Takeuchi, T., \&

1357:     Lin, D. N. C.  2002, \apj, 581, 1344

1358: \bibitem[Takeuchi \& Lin(2003)]{TL03} Takeuchi, T., \&    

1359:     Lin, D. N. C.  2003, \apj, 593, 524

1360: \bibitem[Tanaka et al.(2005)]{THI05} Tanaka, H., Himeno, Y., \&

1361:     Ida, S.  2005, \apj, 625, 414   

1362: \bibitem[Watanabe et al.(1990)]{WNN90} Watanabe, S.,

1363:     Nakagawa, Y., \& Nakazawa, K.  1990, \apj, 358, 282 

1364: \end{thebibliography}

1365: 

1366: \clearpage

1367: 

1368: \begin{figure}

1369: \plotone{f1.ps}

1370: \caption{Disk thermal emission $F\Msub{m}$ ({\it solid line}) and intercepted

1371: stellar fluxes $F\Msub{s}$ at $r = 10 \Punit{AU}$ with constant surface heights 

1372: ({\it dashed curves}) of  $z\Msub{s}/r = 0.12$, $0.13$, and $0.14$ 

1373: ({\it left to right}) and with constant $\chi = z\Msub{s}/h$ ({\it dot-dashed line}) 

1374: as functions of disk temperature $T\Msub{m}$.  All fluxes are normalized 

1375: by $\sigma T_\star^4$.

1376: \label{FIG:FmFs}}

1377: \end{figure}

1378: 

1379: \begin{figure}

1380: \plotone{f2.ps}

1381: \caption{Initial stage of evolution of the midplane temperature

1382: $T\Msub{m}$ for the case of constant opacity and $\beta=0$. The thick

1383: curves cover the range $\hat{t} = t/t\Msub{th,0} = 0.0$--$1.6$ with the

1384: interval $\Delta \hat{t} = 0.2$.  The surface temperature $T\Msub{s}$

1385: is represented by a thin line in the back panel. The surface density

1386: distribution is that prescribed by the MMSN model in which $p=1.5$.

1387: The viscous dissipation associated with mass accretion is neglected

1388: (i.e., $\dot{M} = 0$).

1389: \label{FIG:init-ck-M0-T}}

1390: \end{figure}

1391: 

1392: \begin{figure}

1393: \plotone{f3.ps}

1394: \caption{Time evolution of $T\Msub{m}$ after it has reached a

1395: quasi-periodic state.  This model is the continuation of that shown in

1396: Fig.~\ref{FIG:init-ck-M0-T} to an epoch $\hat{t} = 8.0$--$9.6$.  Each

1397: curve is separated by $\Delta \hat{t} = 0.2$. The $T\Msub{s}$

1398: distribution is also shown in a thin line in the back panel.  Other

1399: parameters are same as Fig.~\ref{FIG:init-ck-M0-T}.

1400: \label{FIG:ck-M0-T}}

1401: \end{figure}

1402: 

1403: \begin{figure}

1404: \plotone{f4.ps}

1405: \caption{Time evolution of $\zeta\Msub{s} = z\Msub{s}/r$ in the

1406: quasi-periodic state at the same epoch as Fig.~\ref{FIG:ck-M0-T}.

1407: \label{FIG:ck-M0-zeta}}

1408: \end{figure}

1409: 

1410: \begin{figure}

1411: \plotone{f5.ps}

1412: \caption{Time evolution of $\chi = z\Msub{s}/h$ in the

1413: quasi-periodic state at the same epoch as Fig.~\ref{FIG:ck-M0-T}.

1414: \label{FIG:ck-M0-zovh}}

1415: \end{figure}

1416: 

1417: \begin{figure}

1418: \plotone{f6.ps}

1419: \caption{Time evolution of surface filling factor $A\Msub{s}$ in

1420: quasi-periodic state at the same epoch as Fig.~\ref{FIG:ck-M0-T}.

1421: \label{FIG:ck-M0-As}}

1422: \end{figure}

1423: 

1424: \begin{figure}

1425: \plotone{f7.ps}

1426: \caption{Time evolution of the logarithmic pressure gradient 

1427: $d \ln P/d \ln r$ in quasi-periodic state at the same epoch as

1428: Fig.~\ref{FIG:ck-M0-T}.

1429: \label{FIG:ck-M0-dP}}

1430: \end{figure}

1431: 

1432: \begin{figure}

1433: \plotone{f8.ps}

1434: \caption{Time evolution of potential vorticity $\Omega\Msub{ep}

1435: \Sigma^{-1}$ normalized by the Keplerian value at $1 \Punit{AU}$

1436: ($\Omega\Msub{K0} \Sigma_0^{-1}$) in quasi-periodic state at the same

1437: epoch as Fig.~\ref{FIG:ck-M0-T}.

1438: \label{FIG:ck-M0-Pv}}

1439: \end{figure}

1440: 

1441: \begin{figure}

1442: \plotone{f9.ps}

1443: \caption{Quasi-periodic evolution of $T\Msub{m}$ for steady

1444: accretion disk with $\dot{M} = 10^{-8} M_\sun \Punit{yr}^{-1}$ (at

1445: $\hat{t} = 14.4$--$16.0$ with $\Delta \hat{t} = 0.2$).  The

1446: distribution of $T\Msub{s}$ is also shown with a thin solid line in

1447: the back panel.  Other parameters are same as

1448: Fig.~\ref{FIG:ck-M0-T}.

1449: \label{FIG:ck-M8-T}}

1450: \end{figure}

1451: 

1452: \begin{figure}

1453: \plotone{f10.ps}

1454: \caption{Quasi-periodic evolution of $T\Msub{m}$ for steady

1455: accretion disk with $\dot{M} = 10^{-7} M_\sun \Punit{yr}^{-1}$ (

1456: $\hat{t} = 11.2$--$14.4$ with $\Delta \hat{t} = 0.4$).  The

1457: distribution $T\Msub{s}$ is also shown with a thin solid line in the

1458: back panel.  Other parameters are same as Fig.~\ref{FIG:ck-M0-T}.

1459: \label{FIG:ck-M7-T}}

1460: \end{figure}

1461: 

1462: \begin{figure}

1463: \plotone{f11.ps}

1464: \caption{Quasi-periodic evolution of $T\Msub{m}$ for a steady

1465: accretion disk with $p = 1.0$ and $\dot{M} = 10^{-8} M_\sun

1466: \Punit{yr}^{-1}$ ($\hat{t} = 9.6$--$12.8$ with $\Delta \hat{t} =

1467: 0.4$).  The distribution of $T\Msub{s}$ is also shown with a thin

1468: solid line in the back panel.  Opacities are same as

1469: Fig.~\ref{FIG:ck-M0-T}.

1470: \label{FIG:ck-M8-p10-T}}

1471: \end{figure}

1472: 

1473: \begin{figure}

1474: \plotone{f12.ps}

1475: \caption{Quasi-periodic evolution of $T\Msub{m}$ for realistic

1476: opacities as well as $T\Msub{s}$ (a thin solid line in the back

1477: panel).  The disk structural parameters is that of a MMSN model

1478: ($p=1.5$) with steady $\dot{M} = 10^{-8} M_\sun \Punit{yr}^{-1}$. The

1479: opacity law changes at $160 \Punit{K}$ but the effect of ice

1480: sublimation is neglected. Maximum size of surface dust grains is

1481: assumed to be $1 \Punit{mm}$ (same as that in disk interior). Times

1482: are same as Fig.~\ref{FIG:ck-M0-T}.

1483: \label{FIG:vk-M8-T}}

1484: \end{figure}

1485: 

1486: \begin{figure}

1487: \plotone{f13.ps}

1488: \caption{Same as Fig.~\ref{FIG:vk-M8-T}, but for a disk with

1489: $\dot{M} = 10^{-7} M_\sun \Punit{yr}^{-1}$ and $\hat{t} = 8.0$--$11.2$

1490: with $\Delta \hat{t} = 0.4$.

1491: \label{FIG:vk-M7-T}}

1492: \end{figure}

1493: 

1494: \begin{figure}

1495: \plotone{f14.ps}

1496: \caption{Same as Fig.~\ref{FIG:vk-M8-T}, but taking into account the

1497: effect of ice sublimation.  The maximum size of surface dust grains is

1498: reduced to $1 \Punit{\micron}$.

1499: \label{FIG:vk-ev-M8-T}}

1500: \end{figure}

1501: 

1502: \begin{figure}

1503: \plotone{f15.ps}

1504: \caption{Quasi-periodic change of SED calculated from the model

1505: shown in Fig.~\ref{FIG:vk-ev-M8-T}.  Different lines represent the

1506: epochs $\hat{t} = 8.4$ ({\it thick solid curve}), $8.6$ ({\it thin solid curve}),

1507: $8.8$ ({\it dashed curve}), $9.0$ ({\it dash-dotted curve}), 

1508: $9.2$ ({\it dotted curve}), and $9.4$ ({\it thick solid curve}, coincide with 

1509: that at $\hat{t}=8.4$), respectively.  Note that $t = 53 \hat{t} \Punit{yr}$.

1510: \label{FIG:vk-ev-M8-SED}}

1511: \end{figure}

1512: 

1513: \end{document}

1514: 

1515: %%

1516: %% End of file `mssei.tex'.

1517: