1: %% This is a manuscript marked up using the

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

3: 

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

5: 

6: %% \documentclass[manuscript]{aastex}

7: % \documentclass[preprint2]{aastex}

8: %% \documentclass[preprint2,longabstract]{aastex}

9: \documentclass[numberedappendix]{emulateapj}

10: \usepackage{amssymb}

11: 

12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

13: \def\deg {{\hbox{$^\circ$}}}

14: \def\eq#1{\begin{equation} #1 \end{equation}}

15: \let\mic=\micron

16: \def\x      {\hbox{$\times$}}

17: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

18: 

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

20: %\slugcomment{Submitted 2005 June 07}

21: 

22: \shorttitle{Temperature inversion in multigrain dust}

23: \shortauthors{Vinkovi\'{c} D.}

24: 

25: \begin{document}

26: 

27: \title{Temperature inversion on the surface of externally heated

28: optically thick multigrain dust clouds}

29: 

30: 

31: \author{Dejan Vinkovi\'{c}}

32: \affil{Institute for Advanced Study, School of Natural Sciences,

33:        Einstein Drive, Princeton, NJ 08540; dejan@ias.edu}

34: 

35: \begin{abstract}

36: It was recently discovered that the temperature in the surface layer

37: of externally heated optically thick gray dust clouds increases with

38: the optical depth for some distance from the surface, as opposed to

39: the normal decrease in temperature with distance in the rest of the

40: cloud. This temperature inversion is a result of efficient absorption

41: of diffuse flux from the cloud interior by the surface dust exposed to

42: the external radiation.  A micron or bigger size grains experience

43: this effect when the external flux is of stellar spectrum. We explore

44: what happens to the effect when dust is a mixture of grain sizes

45: (multigrain).  Two possible boundary conditions are considered: i) a

46: constant external flux without constrains on the dust temperature, and

47: ii) the maximum dust temperature set to the sublimation

48: temperature. We find that the first condition allows small grains to

49: completely suppress the temperature inversion of big grains if the

50: overall opacity is dominated by small grains. The second condition

51: enables big grains to maintain the inversion even when they are a

52: minor contributor to the opacity. In reality, the choice of boundary

53: condition depends on the dust dynamics. When applied to the physics of

54: protoplanetary disks, the temperature inversion leads to a previously

55: unrecognized disk structure where optically thin dust can exist inside

56: the dust destruction radius of an optically thick disk. We conclude

57: that the transition between the dusty disk and the gaseous inner

58: clearing is not a sharp edge, but rather a large optically thin

59: region.

60: \end{abstract}

61: 

62: \keywords{ accretion, accretion disks --- circumstellar matter --- dust,

63: extinction --- stars: pre-main-sequence}

64: 

65: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

66: \section{Introduction}

67: 

68: Dust is one of the principal components of interstellar and

69: circumstellar matter. It serves as a very efficient absorber of

70: starlight, which is dominated by visual and ultraviolet photons, and

71: reemiter of this absorbed energy into infrared (IR), thereby modifying

72: the entire spectral energy distribution. Hence, radiative transfer in

73: dusty environments is an inherent part of the study of star formation,

74: protoplanetary disk evolution, dusty winds from AGB stars,

75: circumnuclear environment of AGNs, etc.

76: 

77: In general, radiative transfer is extremely complicated because the

78: dust in these environments always comes as a mixture of various grain

79: sizes (multigrain\footnote{Typically, the term {\it multigrain} also

80: includes all other dust grain properties, like the grain shape and

81: chemistry. But, for simplicity, we use this term only to describe

82: grain size effects.}), shapes and chemical compositions. It is,

83: therefore, common to employ certain approximations that make the

84: problem computationally manageable. For example, dust grains are

85: usually approximated with spheres of similar chemical composition, so

86: that the Mie theory can be easily used for calculating dust cross

87: sections. Another most commonly used approximation is replacing a

88: mixture of dust grain sizes by an equivalent (i.e. synthetic or

89: average) single grain size.  Numerical calculations have shown that

90: this approximation does not produce a significant change in the

91: spectral energy distribution \citep{ERR94,Wolf,Carciofi}. It is

92: assumed that this is particularly appropriate for optically thick dust

93: clouds.

94: 

95: It has been discovered recently that externally heated optically thick

96: clouds made of large ($\ga 1\mu$m) single size dust grains produce the

97: effect of {\it temperature inversion} where the maximum temperature is

98: within the dust cloud\footnote{Another type of temperature inversion

99: has been recognized in protoplanetary disks, where additional viscous

100: heating can increase the disk interior temperature

101: \citep{Calvet,Malbet}. In contrast, the temperature inversion

102: discussed here is a pure radiative transfer effect and does not

103: require any additional assumption (like disk viscosity) to operate.},

104: at the visual optical depth of $\tau_V\sim$1, instead of on the very

105: surface exposed to the stellar heating \citep{Isella,VIJE06}. The

106: cause of this inversion within the surface layer is its ability to

107: efficiently absorb the diffuse IR radiation originating from the

108: cloud's interior - a process similar to the ``greenhouse effect.''

109: 

110: It is not known, however, if multigrain dust would produce the same

111: temperature inversion effect. Here we employ analytical multigrain

112: radiative transfer to study the surface of optically thick clouds.

113: Two possible types of boundary conditions are explored: i) a constant

114: external flux heating the cloud, with no limits on the dust

115: temperature, and ii) a fixed maximum dust temperature corresponding to

116: dust sublimation. In \S\ref{single_grain} we describe the analytic

117: method and apply it on single size dust. In the next section

118: \S\ref{multigrain} we apply the method on multigrain dust. After that

119: in \S\ref{transverse} we explore a possibility of surface thermal

120: cooling in transverse direction.  Some aspects of our result are

121: discussed in \S\ref{discussion}, with our conclusion in

122: \S\ref{conclusion}.

123: 

124: \section{Optically thick single size grain dust cloud}

125: \label{single_grain}

126: 

127: In order to understand effects of multigrain dust on the temperature

128: structure of optically thick clouds, we first have to take a look at

129: the single size dust grain clouds. Here we reiterate analytic

130: techniques described in the literature \citep{Isella,VIJE06} and then

131: expand our approach to multigrain dust in the following sections.

132: 

133: Consider dust grains at three different locations (Figure \ref{Sketch1}):

134: on the very surface (point {\bf P$_0$}), at optical depth $\tau_V$

135: (point {\bf P$_1$}) and at optical depth $2\tau_V$ (point {\bf

136: P$_2$}). The optical depth $\tau_V$ is defined at the peak wavelength

137: of the bolometric temperature of external flux $F_{in}$ illuminating

138: the cloud. We will be interested in cases where dust can reach

139: sublimation temperatures (between $\sim$1,000K and 2,000K for

140: interstellar dust) and where the external flux source is a star-like

141: object peaking its emission at submicron wavelengths. For the purpose

142: of this paper we use the optical depth $\tau_V$ in visual, but one can

143: adjust it to the required wavelength of interest.

144: 

145: The surface layer between points {\bf P$_0$} and {\bf P$_1$} emits the

146: thermal flux $F_{sur1}$. We assume that it emits equally on both

147: sides (toward the left and right in Figure \ref{Sketch1}). This

148: approximation is valid if the layer is optically thin at the peak

149: wavelength of its IR emission. If we define $q=\sigma_V/\sigma_{IR}$,

150: where $\sigma_V$ and $\sigma_{IR}$ are dust absorption cross sections in

151: visual and IR, then the IR optical depth requirement is

152:  \eq{\label{tauIR}

153:    \tau_{IR}={\tau_V\over q}\la 1.

154:  }

155: The same requirement holds for the second surface layer between points

156: {\bf P$_1$} and {\bf P$_2$} and its thermal flux $F_{sur2}$.

157: Each of these two layers attenuates the external flux by

158: $exp(-\tau_V)$ and the thermal fluxes by

159: $exp(-\tau_{IR})=exp(-\tau_V/q)$.

160: 

161: Since the cloud is optically thick to its own radiation, no net

162: flux can go through the cloud. The total flux coming from the left

163: at any point in the cloud is equal to the total flux coming from

164: the right. This yields flux balance equations at points {\bf

165: P$_0$}, {\bf P$_1$} and {\bf P$_2$}

166:  \eq{\label{F0}

167:   F_{in} = F_{sur1} + F_{sur2}\, e^{-\tau_V/q} + F_{out}\, e^{-2\tau_V/q}

168:  }

169:  \eq{\label{F1}

170:   F_{in}\, e^{-\tau_V}+ F_{sur1} = F_{sur2} + F_{out}\, e^{-\tau_V/q}

171:  }

172:  \eq{\label{F2}

173:   F_{in}\, e^{-2\tau_V} + F_{sur1}\, e^{-\tau_V/q} + F_{sur2} = F_{out}

174:  }

175: where $F_{out}$ is the thermal flux coming out of the cloud interior

176: at {\bf P$_2$}.

177: 

178: We want the external flux $F_{in}$ at point {\bf P$_2$} to be attenuated enough

179: to make diffuse flux the dominant source of heating.

180: This requirement means that $2\tau_V\ga 1$, which in combination with

181: equation \ref{tauIR} gives the allowed range for $\tau_V$

182:  \eq{

183:    0.5\la \tau_V \la q.

184:  }

185: From equations \ref{F0}-\ref{F2} we further get

186:  \eq{\label{Fsur2}

187:   F_{sur2}=F_{in} {1+e^{-\tau_V} \over 1+e^{-\tau_V/q}} - F_{out}e^{-\tau_V/q}

188:  }

189:  \eq{\label{Fsur1}

190:   F_{sur1}=F_{sur2} - F_{in} e^{-\tau_V} {1+e^{-\tau_V} \over 1+e^{-\tau_V/q}}.

191:  }

192: 

193: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

194: \begin{figure}

195: \epsscale{1.15}

196:  \plotone{Figure1.ps}

197:  \caption{\label{Sketch1} Sketch of the surface of an optically thick

198: single size grain dust cloud illuminated from outside.  See

199: \S\ref{single_grain} for details.}

200: \end{figure}

201: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

202: 

203: A dust grain at {\bf P$_0$} absorbs $\sim \sigma_V F_{in}$ of the

204: external directional flux. It also absorbs $\sim 2\sigma_{IR}F_{IR}$

205: of any infrared diffuse flux $F_{IR}$, where the factor 2 accounts

206: for absorption from 2$\pi$ sr. The grain emits at its temperature

207: $T_0$ into 4$\pi$ sr, so that the energy balance is

208:  \[

209:   \sigma_VF_{in} + 2\sigma_{IR}(F_{sur1} + F_{sur2}\, e^{-\tau_V/q}

210:      + F_{out}\, e^{-2\tau_V/q}) =

211:  \]

212:  \eq{

213:      = 4\sigma_{IR}\sigma_{SB}T_0^4

214:  }

215: where $\sigma_{SB}$ is the Stefan-Boltzmann constant. Using the flux balance

216: in equation \ref{F0} and $q=\sigma_V/\sigma_{IR}$, we get

217:  \eq{\label{Fin}

218:    F_{in}={4\sigma_{SB}\over q+2} T_0^4.

219:  }

220: 

221: Similarly, we can write the energy balance for a dust grain at {\bf

222: P$_1$} and {\bf P$_2$}

223:  \eq{\label{ebT1}

224:   qF_{in}\, e^{-\tau_V}+ 2F_{sur1} + 2F_{sur2} + 2F_{out}\, e^{-\tau_V/q}

225:     = 4\sigma_{SB}T_1^4

226:  }

227:  \eq{\label{ebT2}

228:   qF_{in}\, e^{-2\tau_V} + 2F_{sur1}\, e^{-\tau_V/q} + 2F_{sur2} + 2F_{out}.

229:     = 4\sigma_{SB}T_2^4

230:  }

231: 

232: We approximate the interior flux as

233:  \eq{\label{Fout}

234:   F_{out} \sim \sigma_{SB}T_2^4.

235:  }

236: This is a good approximation of the interior for gray dust and an

237: overestimate for non-gray dust, where temperature decreases

238: with optical depth.

239: Now we can continue deriving temperatures in two possible ways.

240: 

241: 

242: {\it METHOD 1}: Combining equations \ref{Fsur2},\ref{Fsur1},

243: \ref{Fin}, \ref{ebT1}, \ref{ebT2} and \ref{Fout} yields

244:  \eq{\label{T1_method1}

245:   T_1^4 = {A+B \over q+2} T_0^4

246:  }

247:  \eq{\label{T2_method1}

248:   T_2^4 = {2B \over q+2} T_0^4

249:  }

250:  \eq{\label{A_method1}

251:   A = qe^{-\tau_V}+2(1-e^{-\tau_V}){1+e^{-\tau_V}\over 1+e^{-\tau_V/q}}

252:  }

253:  \eq{\label{B_method1}

254:   B = qe^{-2\tau_V}+2(1-e^{-\tau_V-\tau_V/q}){1+e^{-\tau_V}\over 1+e^{-\tau_V/q}}.

255:  }

256: 

257: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

258: \begin{figure}

259: \epsscale{1.15}

260:  \plotone{Figure2.ps}

261:  \caption{\label{Temperature1}

262: Surface temperature of an optically thick single size grain dust cloud for various

263: visual to infrared opacity ratios $q$.

264: The upper panel shows the solution based on equations \ref{T1_method1} and \ref{T2_method1}

265: (method 1). The lower panel is based on equations \ref{T2_method2} and \ref{T1_method2}

266: (method 2). Optical depth step is $\tau_V=1$. Both solutions show temperature inversion

267: (increase of temperature with optical depth) for small $q$.

268:  }

269: \end{figure}

270: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

271: 

272: The upper panel of Figure \ref{Temperature1} shows this temperature profile for $\tau_V=1$. The

273: effect of temperature inversion $T_2>T_0$ is present when

274: \hbox{$q<q_{limit}\sim 1.4$}. Only grains larger than about one micron

275: can have such a small value of $q$, which makes them almost ``gray''

276: in the near IR. This inversion does not appear in non-gray dust

277: ($q>q_{limit}$) where the temperature decreases monotonically with

278: distance from the cloud surface.

279: 

280: 

281: {\it METHOD 2}: We can use equations \ref{F0}-\ref{F2} to express

282: $F_{out}$ as a function of $F_{in}$ and then use equations \ref{Fin} and

283: \ref{Fout} to obtain

284:  \eq{\label{T2_method2}

285:   T_2^4 = {4\over q+2}\cdot {1+e^{-\tau_V}(1+e^{-\tau_V}-e^{-\tau_V/q}).

286:            \over 1+e^{-\tau_V/q}} T_0^4

287:  }

288: Similarly, we can express $F_{sur1}$ and $F_{sur2}$ as a function of $F_{in}$

289: and then use equation \ref{ebT1} to obtain

290:  \eq{\label{T1_method2}

291:   T_1^4 = {1\over q+2}\left[ qe^{-\tau_V} +

292:            2{2+e^{-\tau_V}-e^{-\tau_V - \tau_V/q} \over 1+e^{-\tau_V/q}} \right]T_0^4.

293:  }

294: The lower panel of Figure \ref{Temperature1} shows this result for $\tau_V=1$.

295: It differs slightly from previous solution in Method 1, but gives qualitatively the same

296: result. The temperature inversion now exists for

297: \hbox{$q<q_{limit}\sim 1.5$}. Henceforth, we use Method 1 for further analysis.

298: 

299: \section{Optically thick multigrain cloud}

300: \label{multigrain}

301: 

302: \subsection{Constant external flux as the boundary condition}

303: \label{multigrain_const_flux}

304: 

305: We consider now an externally heated optically thick dust cloud made

306: of $N$ grain sizes, under the condition of constant external flux and

307: no limits on the dust temperature. A dust grain of the $i^{th}$ size

308: at a point of $\tau_V$ optical distance from the cloud surface is

309: heated by the attenuated stellar flux $F_{in}exp(-\tau_V)$, by the

310: local diffuse flux $F_{sur}$ coming from the direction of the surface,

311: and by the local diffuse flux $F_{out}$ coming out of the cloud

312: interior. The thermodynamic equilibrium of the dust grain gives

313:  \eq{\label{ebi}

314:  \sigma_{V,i}\, F_{in}e^{-\tau_V} +2\sigma_{IR,i}\, (F_{sur} +

315:       F_{out})  = 4\sigma_{IR,i}\sigma_{SB}\, T_i^4(\tau_V).

316:  }

317: Factors 2 and 4 account for absorption from 2$\pi$ sr and emission

318: into 4$\pi$ sr.

319: 

320: On the very surface of the cloud $\tau_V=0$ and $F_{sur}=0$, therefore

321: $F_{in}=F_{out}$, which yields

322:  \eq{\label{Fini}

323:     T_i^4(0) = (q_i+2){F_{in}\over 4\sigma_{SB}}

324:  }

325: where $q_i=\sigma_{V,i}/\sigma_{IR,i}$.  This shows that {\it dust

326: grains of different sizes have different temperatures on the surface

327: of an optically thick cloud}.  The lowest temperature among the grain

328: sizes is acquired by the largest grains because they have the smallest

329: $q_i$.

330: 

331: On the other hand, the contribution of the external flux is negligible

332: $F_{in}exp(-\tau_V)\to 0$ when $\tau_V\gg 1$, therefore

333: $F_{sur}=F_{out}$, which yields {\it the same temperature for all grain sizes}

334:  \eq{\label{T_tauV_gg_1}

335:    T_i^4(\tau_V\gg 1) =  F_{out}/\sigma_{SB}.

336:  }

337: 

338: A general relationship between grain temperatures is derived by

339: subtracting equation \ref{ebi} for a grain $i$ from the same equation

340: for a grain $j$ and then use equation \ref{Fini} to obtain

341:  \eq{\label{Tij}

342:   T_j^4(\tau_V) = T_i^4(\tau_V) + {q_j-q_i\over q_i+2}e^{-\tau_V} T_i^4(0)

343:  }

344: Since $q$ scales inversely with the grain size, equation

345: \ref{Tij} shows that {\it smaller grains always have a higher

346: temperature than bigger grains at any point in the cloud}, with

347: the limit $T_j \sim T_i$ when $exp(-\tau_V)\ll 1$.

348: 

349: Optical depth is now a cumulative contribution of all grain sizes in the mix.

350:  \eq{

351:    \tau_\lambda \propto \sum\limits_{i=1}^N n_i\sigma_{\lambda,i},

352:  }

353: where $n_i$ is the number density of the i$^{th}$ grain size.

354: If we scale optical depth relative to $\tau_V$ then

355:  \eq{

356:    \tau_{IR} = \tau_V { \sum_{i=1}^N n_i\sigma_{IR,i}\over

357:                         \sum_{i=1}^N n_i\sigma_{V,i} } =

358:                \tau_V \sum\limits_{i=1}^N {\Upsilon_{V,i}\over q_i},

359:  }

360: where $\Upsilon_{\lambda,i}$ is the relative contribution of the $i^{th}$ grain

361: to the dust opacity at wavelength $\lambda$

362:  \eq{

363:    \Upsilon_{\lambda,i} = { n_i\sigma_{\lambda,i}\over

364:                         \sum_{j=1}^N n_j\sigma_{\lambda,j} }.

365:  }

366: 

367: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

368: \begin{figure}

369: \epsscale{1.15}

370:  \plotone{Figure3.ps}

371:  \caption{\label{Temp_multigrain_whole_fig1}

372: Temperature of two types of grains in the surface of an optically

373: thick multigrain dust cloud: larger grains $\alpha$ ({\it thick

374: lines}, equations \ref{T1alpha_whole} and \ref{T2alpha_whole}) and

375: smaller grains $\beta$ ({\it thin lines}, equation \ref{Tsmall_n}).

376: Line styles correspond to different ratios $q_\alpha$

377: of visual to infrared opacity of big grains. The same ratio for small

378: grains is fixed to $q_\beta=q_\alpha+1$.  Panels show results for two

379: relative contributions of big grains to the visual opacity:

380: $\Upsilon_{V\alpha}=0.1$ in the upper panel and

381: $\Upsilon_{V\alpha}=0.9$ in the lower panel.  Optical depth step is

382: $\tau_V=1$.

383: Notice how the temperature inversion in big grains (small $q_\alpha$)

384: disappears when small grains dominate the overall opacity (upper

385: panel). The inversion exists only if bigger grains dominate (lower

386: panel).

387:  }

388: \end{figure}

389: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

390: 

391: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

392: \begin{figure}

393: \epsscale{1.15}

394:  \plotone{Figure4.ps}

395:  \caption{\label{Temp_multigrain_whole_fig2}

396: Same as Fig. \ref{Temp_multigrain_whole_fig1}, but for $q_\beta=q_\alpha+10$.

397:  }

398: \end{figure}

399: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

400: 

401: Consider now a model similar to figure \ref{Sketch1}, except that the dust

402: is multigrain.  For simplicity and clarity of the following

403: analysis, we work with only two grain sizes: a ``big'' grain {\it

404: $\alpha$} with $q_\alpha\sim 1$ and a ``small'' grain {\it $\beta$}

405: with $q_\beta>q_\alpha$.  The infrared optical depth step is

406:  \eq{\label{tauIR_multigrain}

407:   \tau_{IR}=\tau_V\left({\Upsilon_{V,\alpha}\over q_\alpha} +

408:                         {\Upsilon_{V,\beta}\over q_\beta} \right).

409:  }

410: According to equation \ref{Tij} the temperature of small grains at

411: optical depth $k\tau_V$ (point {\bf P$_k$}) is

412:  \eq{\label{Tsmall_n}

413:   T_{k,\beta}^4 = T_{k,\alpha}^4 + {q_\beta-q_\alpha\over q_\alpha+2}e^{-k\tau_V}

414:                                    T_{0,\alpha}^4.

415:  }

416: The temperature of big grains at point {\bf P$_0$} is (equation \ref{Fini})

417:  \eq{\label{T_0alpha}

418:   T_{0,\alpha}^4 = (q_\alpha+2){F_{in}\over 4\sigma_{SB}},

419:  }

420: while for the other temperatures we need the flux balance at points

421: {\bf P$_0$}, {\bf P$_1$} and {\bf P$_2$}. Since the balance  is

422: the same as in equations \ref{F0}-\ref{F2}, except that the infrared

423: step $\tau_V/q$ is replaced by $\tau_{IR}$, we use the procedure described in

424: \S\ref{single_grain} and derive

425:  \eq{\label{T1alpha_}

426:   T_{1,\alpha}^4 = {A_\alpha \over q_\alpha+2}T_{0,\alpha}^4 +

427:                    {F_{out}\over 2\sigma_{SB}}

428:  }

429:  \eq{\label{T2alpha_}

430:    T_{2,\alpha}^4 = {B_\alpha \over q_\alpha+2}T_{0,\alpha}^4 +

431:                    {F_{out}\over 2\sigma_{SB}}

432:  }

433:  \eq{

434:   A_\alpha = q_\alpha\, e^{-\tau_V}+

435:              2\,(1+e^{-\tau_V}){1+e^{-\tau_V}\over 1+e^{-\tau_{IR}}}

436:  }

437:  \eq{

438:   B_\alpha = q_\alpha\, e^{-2\tau_V}+

439:              2\,(1+e^{-\tau_V-\tau_{IR}}){1+e^{-\tau_V}\over 1+e^{-\tau_{IR}}}.

440:  }

441: 

442: The interior flux $F_{out}$ is now a cumulative contribution of all

443: grain sizes according to their relative contribution to the dust

444: opacity. In our two-size example

445:  \eq{\label{Fout_multi}

446:   F_{out}/\sigma_{SB}=\Upsilon_{IR,\alpha} T_{2,\alpha}^4 +

447:                       \Upsilon_{IR,\beta} T_{2,\beta}^4.

448:  }

449: Combined with equation \ref{Tsmall_n} gives

450:  \eq{\label{Foutc}

451:   F_{out}/\sigma_{SB}=T_{2,\alpha}^4 + {C_\alpha\over q_\alpha+2}T_{0,\alpha}^4

452:  }

453:  \eq{\label{calpha}

454:    C_\alpha = \Upsilon_{IR,\beta}(q_\beta - q_\alpha)e^{-2\tau_V}.

455:  }

456: 

457: Putting together equations \ref{T1alpha_}, \ref{T2alpha_} and

458: \ref{Foutc} yields the solution

459:  \eq{\label{T1alpha_whole}

460:   T_{1,\alpha}^4 = {A_\alpha+ B_\alpha+C_\alpha\over q_\alpha+2}T_{0,\alpha}^4

461:  }

462:  \eq{\label{T2alpha_whole}

463:   T_{2,\alpha}^4 = {2B_\alpha+C_\alpha\over q_\alpha+2}T_{0,\alpha}^4.

464:  }

465: 

466: The resulting temperature is plotted in Figures

467: \ref{Temp_multigrain_whole_fig1} and \ref{Temp_multigrain_whole_fig2}.

468: The upper panel in each figure shows the result when small grains

469: dominate the opacity ($\Upsilon_{V,\alpha}=0.1$). The temperature

470: inversion in big grains is suppressed because the local diffuse flux

471: is dictated by small grains. If big grains dominate the opacity (lower

472: panels, $\Upsilon_{V,\alpha}=0.9$) then the temperature inversion is

473: preserved.

474: 

475: 

476: \subsection{Sublimation temperature as the boundary condition}

477: \label{multigrain2}

478: 

479: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

480: \begin{figure}

481: \epsscale{1.15}

482:  \plotone{Figure5.ps}

483:  \caption{\label{Sketch2}

484: Sketch of the surface of an optically thick multigrain dust cloud illuminated from outside.

485:  See \S\ref{multigrain2} for details.

486:  }

487: \end{figure}

488: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

489: 

490: Now we consider a dust cloud hot enough on its illuminated surface to

491: sublimate dust grains warmer than the sublimation temperature

492: $T_{sub}$. The flux entering the cloud is adjustable to accommodate

493: any temperature boundary condition. From equation \ref{Fini} we see

494: that small grains are the first to be removed from the immediate

495: surface. If the external flux is high enough then the immediate

496: surface is populated only by $q_i\sim 1$ grains (``big grains''). All

497: other grains (``small grains'') would survive somewhere within the

498: cloud, at a distance where the local flux is reddened enough by big

499: grains to be absorbed less efficiently.

500: 

501: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

502: \begin{figure}

503: \epsscale{1.15}

504:  \plotone{Figure6.ps}

505:  \caption{\label{Temp_multigrain_fig1}

506: Same as Figure \ref{Temp_multigrain_whole_fig1}, except that small

507: grains are removed from the immediate surface by sublimation (Figure

508: \ref{Sketch2}). The temperature of larger grains is calculated from

509: equations \ref{T1alpha} and \ref{T2alpha}.  Despite being mixed with

510: smaller grains, larger grains show temperature inversion even when

511: smaller grains dominate the opacity.

512: }

513: \end{figure}

514: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

515: 

516: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

517: \begin{figure}

518: \epsscale{1.15}

519:  \plotone{Figure7.ps}

520:  \caption{\label{Temp_multigrain_fig2}

521: Same as Figure \ref{Temp_multigrain_fig1}, but for $q_\beta=q_\alpha+10$.

522:  }

523: \end{figure}

524: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

525: 

526: We again apply our two-size dust model, except that the surface layer

527: of visual optical depth $\tau_V$ is occupied only by big grains (see

528: Figure \ref{Sketch2}). It is too hot for small grains to survive

529: within this layer. Both grains exist at optical distances larger than

530: $\tau_V$ from the surface.

531: Following the same procedure as in \S\ref{single_grain}, we write

532: the flux balance at points {\bf P$_0$}, {\bf P$_1$} and {\bf P$_2$} (see

533: Figure \ref{Sketch2})

534:  \eq{\label{F0multi}

535:   F_{in} = F_{sur1} + F_{sur2}\, e^{-\tau_V/q_\alpha} + F_{out}\, e^{-\tau_V/q_\alpha-\tau_{IR}}

536:  }

537:  \eq{\label{F1multi}

538:   F_{in}\, e^{-\tau_V}+ F_{sur1} = F_{sur2} + F_{out}\, e^{-\tau_{IR}}

539:  }

540:  \eq{\label{F2multi}

541:   F_{in}\, e^{-2\tau_V} + F_{sur1}\, e^{-\tau_{IR}} + F_{sur2} = F_{out}

542:  }

543: where the IR optical depth between points {\bf P$_1$} and {\bf P$_2$}

544: is given in equation \ref{tauIR_multigrain}.

545: 

546: Small grains do not exist now at point {\bf P$_0$}, but their

547: temperature at other points is still described by equation

548: \ref{Tsmall_n}.  Deriving big grain temperatures at {\bf

549: P$_1$} and {\bf P$_2$} is now a straightforward procedure already

550: described in \S\ref{single_grain} and \S\ref{multigrain_const_flux}

551:  \eq{

552:   A_\alpha^\prime = q_\alpha\, e^{-\tau_V}+2\, {1+e^{-\tau_V}\over 1+e^{-\tau_V/q_\alpha}}

553:                    -2\, e^{-\tau_V}{1+e^{-\tau_V}\over 1+e^{-\tau_{IR}}}.

554:  }

555:  \eq{

556:   B_\alpha^\prime = q_\alpha\, e^{-2\tau_V}+2\, {1+e^{-\tau_V}\over 1+e^{-\tau_V/q_\alpha}}

557:                    -2\, e^{-\tau_V-\tau_{IR}}{1+e^{-\tau_V}\over 1+e^{-\tau_{IR}}}

558:  }

559:  \eq{\label{T1alpha}

560:   T_{1,\alpha}^4 = {A_\alpha^\prime+ B_\alpha^\prime+C_\alpha\over q_\alpha+2}T_{0,\alpha}^4

561:  }

562:  \eq{\label{T2alpha}

563:   T_{2,\alpha}^4 = {2B_\alpha^\prime+C_\alpha\over q_\alpha+2}T_{0,\alpha}^4.

564:  }

565: The resulting temperature is plotted in Figures \ref{Temp_multigrain_fig1}

566: and \ref{Temp_multigrain_fig2}. Two

567: important results are deduced for big grains from this solution:

568: i) big grains maintain the temperature inversion and ii) if small

569: grains set the maximum temperature limit then big grains, which are always

570: colder than small grains (see equation \ref{Tij}), cannot reach

571: the sublimation temperature.

572: 

573: However, since we do not put limits on the external flux, the

574: solution that allows the maximum possible external flux is the one

575: that also maximizes $F_{out}$. From equation \ref{Fout_multi} we see

576: that the maximum $F_{out}=\sigma_{SB}T_{sub}$ is achieved when

577: $T_{2,\alpha}=T_{2,\beta}=T_{sub}$. Such a solution is not possible

578: unless $\Upsilon_{IR,\beta}=0$. Therefore, {\it dust temperatures

579: are maximized when all small grains sublimate away from the

580: surface region and exist only at optical depths of

581: $exp(-\tau_V)\ll 1$. In other words, the external flux is

582: maximized when the cloud surface "belongs" exclusively to big

583: grains only}.

584: 

585: \section{Transverse cooling of the cloud surface}

586: \label{transverse}

587: 

588: Solutions presented so far assume no time variability in any of the

589: model parameters. The external flux is adjusted by hand to the value

590: that maximizes the external flux and keeps dust temperatures below

591: sublimation. In reality, however, this is a dynamical process where the

592: equilibrium is established by dust moving around and sublimating

593: whenever its temperature exceeds the sublimation point.

594: 

595: The existence of temperature inversion is difficult to understand

596: under such dynamical conditions. Since the very surface of the

597: cloud is {\it below} the sublimation point, its dust can survive

598: closer to the external energy source than the dust within the

599: cloud. Therefore, the distance of the cloud from the source is not

600: defined by the very surface, but rather by the dust at the peak

601: temperature {\it within} the cloud. Imagine now that the whole

602: cloud is moving closer to the source. From the cloud point of

603: reference, this dynamical process is equivalent to increasing the

604: external flux. According to solutions in \S\ref{single_grain} and

605: \S\ref{multigrain}, the peak temperature within the cloud will

606: exceed sublimation at a certain distance from the source and the

607: dust will start to sublimate. This point is the distance at which

608: the dust cloud seemingly stops. However, there is nothing to stop

609: the very surface layer of the cloud to move even closer than the

610: rest of the cloud because its temperature is below sublimation.

611: 

612: Notice that we can not resolve this issue in the approximation of an

613: infinite dusty slab because the transverse optical depth (parallel to

614: the slab surface) is always infinite. No flux can escape the slab in

615: the transverse direction. Hence, the radiative transfer in a slab does

616: not depend on spatial scale and only the optical depth matters. The

617: spatial extension of the surface layer is irrelevant and solutions

618: from \S\ref{single_grain} and \S\ref{multigrain} are applicable

619: irrespectively of the dust dynamics on the spatial scale.

620: 

621: In reality, however, the cloud is finite and dust sublimation can

622: eventually make the cloud optically thin in the transverse

623: direction. This optical depth gap enables thermal radiation to flow

624: out in the transverse direction and thus provides a channel for

625: thermal cooling. Under such conditions the surface layer cam move

626: closer and closer to the energy source by simultaneously expanding the

627: size of the transverse optical depth gap, which then increases the

628: amount of escaping thermal flux. Since the very surface dust is

629: getting closer to the energy source, its temperature increases and

630: eventually it reaches the dust sublimation. At that moment the

631: radiative and dynamical equilibria are established and the whole dust

632: cloud seemingly stops.

633: 

634: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

635: \begin{figure}

636: \epsscale{1.15}

637:  \plotone{Figure8.ps}

638:  \caption{\label{Sketch3}

639: Sketch of the surface of an optically thick dust cloud illuminated from outside and

640: cooled through the surface layer in the transverse direction.

641:  See \S\ref{transverse} for details.

642:  }

643: \end{figure}

644: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

645: 

646: The resulting dust cloud has a large surface zone where the

647: thermal flux can transversely escape the cloud. Here we make an

648: attempt to describe this on a quantitative level through a model

649: shown in Figure \ref{Sketch3}. It is similar to the infinite slab model

650: described in \S\ref{single_grain} except that now we introduce the

651: flux $F_{exit}$ exiting the cloud transversely through the surface

652: layer.

653: 

654: The flux balance equations \ref{F1} and \ref{F2} remain unchanged,

655: while equation \ref{F0} is changed to

656:  \eq{\label{F0_exit}

657:   F_{in} = F_{exit} + F_{sur1} + F_{sur2}\, e^{-\tau_V/q} + F_{out}\,

658:   e^{-2\tau_V/q}.

659:  }

660: By following the same procedure as in \S\ref{single_grain}, we

661: obtain temperatures $T^{\prime\prime}_k$ at points {\bf P$_k$}

662:  \eq{\label{Fin_thin}

663:    F_{in}={4\sigma_{SB}\over q+2} T^{\prime\prime 4}_0 +{2\over q+2} F_{exit}

664:  }

665: \eq{\label{T1_thin}

666:   T^{\prime\prime 4}_1 = {A+B \over q+2} T^{\prime\prime 4}_0 -\left({1\over 1+e^{-\tau_V/q}}-{A+B\over 2(q+2)}\right)F_{exit}

667:  }

668:  \eq{\label{T2_thin}

669:   T^{\prime\prime 4}_2 = {2B \over q+2} T^{\prime\prime 4}_0 -\left({1\over 1+e^{-\tau_V/q}}-{B\over q+2}\right)F_{exit}

670:  }

671: where A and B are given in equations \ref{A_method1} and

672: \ref{B_method1}. The values in parentheses are always positive.

673: 

674: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

675: \begin{figure}

676: \epsscale{1.15}

677:  \plotone{Figure9.ps}

678:  \caption{\label{Temperature_exit}

679: Surface temperature of an optically thick single size grain dust cloud

680: cooled transversely through the surface layer (equations \ref{T1_thin}

681: and \ref{T2_thin}).  The amount of transverse cooling is adjusted to

682: the maximum temperature $T(0)$ at the very surface (equation

683: \ref{Fexit_min}).  Line styles correspond to different ratios $q$ of

684: visual to infrared opacity.  Optical depth step is $\tau_V=1$.

685: Increase in $T(0)$, combined with the temperature inversion (see also

686: Fig. \ref{Temperature1}), creates a local temperature minimum within

687: the surface layer.

688:  }

689: \end{figure}

690: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

691: 

692: We compare this with the original solution in equations \ref{Fin},

693: \ref{T1_method1} and \ref{T2_method1}. Notice that keeping the

694: interior temperature to be the same as before

695: ($T^{\prime\prime}_2=T_2$) requires a larger surface temperature

696: ($T^{\prime\prime}_0>T_0$). If the surface dust of

697: $T^{\prime\prime}_0$ is dynamically driven toward the external

698: energy source, as we discussed above, then it will eventually

699: reach the sublimation temperature $T_{sub}$. This dynamical

700: process is accompanied by dust sublimation in the cloud interior.

701: Sublimation creates a transverse optical depth gap and maintains

702: $T^{\prime\prime}_2=T_{sub}$. The established equilibrium

703: condition $T^{\prime\prime}_0=T^{\prime\prime}_2=T_{sub}$ yields

704:  \eq{\label{Fexit_min}

705:  F_{exit} = {2B-q-2 \over

706:  (q+2)/(1+e^{-\tau_V/q})-B}\,\,\,T^4_{sub}.

707:   }

708: This flux is positive for values of $q$ that produce temperature

709: inversion ($q<q_{limit}$). It originates from the

710: diffuse radiation trapped within the cloud. 

711: Figure \ref{Temperature_exit} shows temperature profiles for

712: $q<q_{limit}$ based on $F_{exit}$ from equation \ref{Fexit_min}.

713: The surface layer now has a local temperature minimum at

714: $T^{\prime\prime}_1$.

715: 

716: \section{Discussion}

717: \label{discussion}

718: 

719: According to equation \ref{T_tauV_gg_1}, the temperature of dust

720: optically deep inside an optically thick cloud depends solely on

721: the local diffuse flux. Since this can give a misleading

722: impression that temperature effects on the cloud surface have no

723: influence on the cloud interior, it is worth explaining why the

724: surface solution is so important for the overall solution.

725: 

726: At optical depths $\tau_V\ga 2$ the flux is exclusively diffuse and

727: the dust temperature is dictated by equation

728: \ref{T_tauV_gg_1}. Notice, however, that even though we set the limit

729: of zero net flux flowing through the cloud, there is no limit on the

730: absolute value of the local diffuse flux. Arbitrarily large but equal

731: diffuse fluxes can flow in opposite directions producing the net zero

732: flux. It is the radiative transfer in the surface layer that sets the

733: absolute scale. The cloud surface of $\tau_V\la 2$ converts the

734: directional external flux of stellar spectral shape into the diffuse

735: thermal flux. Hence, this conversion determines the value of diffuse

736: flux at $\tau_V\sim 2$, which then propagates into the cloud interior

737: of $\tau_V\ga 2$.

738: 

739: 

740: A more rigorous analytic approach shows that the net flux is not

741: exactly zero. A very small flux, in comparison with the external flux,

742: goes through the cloud and creates a temperature gradient. This yields

743: the well known gray opacity solution \citep{Mihalas} where

744: $T^4(\tau)=a\tau+b$. While the constant $a$ depends only on the net

745: flux ($a=0$ in the case of zero net flux), the constant $b$ is

746: determined by the boundary condition on the cloud surface. Hence, this

747: becomes a reiteration of the role of the surface layer described

748: above.

749: 

750: The possibility of the temperature inversion effect described in this

751: paper was discussed previously by \citet{Wolf}. He noticed in his

752: numerical calculations that larger grains can have a higher

753: temperature than smaller grains in the surface of circumstellar dusty

754: disks. He correctly attributed this effect to the more efficient

755: heating of large grains by the IR radiation from the disk interior,

756: but did not analyze it further. This is the same effect noticed

757: by \citet{D02} in his numerical models of circumstellar disks with

758: gray dust opacities. However, it was not until \citet{Isella} and

759: \citet{VIJE06} described this effect analytically in more detail that

760: its importance to the overall disk structure was finally recognized.

761: 

762: The main driver for the detailed description of the temperature

763: inversion effect came from the need for a better understanding of

764: inner regions of dusty disks around young pre-main-sequence stars.

765: The advancements of near infrared interferometry enabled direct

766: imaging of these inner disk regions and resulted in the discovery of

767: inner disk holes produced by dust sublimation \citep{interfer}.

768: However, developing a self-consistent model that would incorporate the

769: spectral and interferometric data proved to be a difficult

770: problem. Two competing models are proposed. In one the data are

771: explained by a disk that has a large vertical expansion (puffing) at

772: the inner dust sublimation edge due to the direct stellar heating of

773: the disk interior \citep{DDN}. In the other model the inner disk is

774: surrounded by an optically thin dusty outflow, without a need for

775: special distortions to the vertical disk structure \citep{VIJE06}.

776: 

777: Although current observations cannot distinguish between these two models,

778: theory gives some limits on the dust properties in the former model.

779: The inner disk edge has to be populated by big dust grains ($\ga

780: 1\mic$) \citep{Isella, VIJE06} in order to produce a large and

781: bright vertical disk puffing needed to fit the data. The hallmark

782: of this radiative transfer solution is the dust temperature

783: inversion, which we also demonstrate in \S\ref{single_grain}.

784: \citet{VIJE06} argue that purely big grains are

785: unrealistic because dust always comes as a mixture of grain sizes

786: (especially in dusty disks where dust collisions should constantly

787: keep small grains in the mix). On the other hand, \citet{Isella}

788: point out that smaller grains should be sublimated away from the

789: very surface of the inner disk edge, but cannot prove that this

790: process is efficient enough to preserve dust temperature

791: inversion.

792: 

793: Our analytical analysis in \S\ref{multigrain} proves that multigrain

794: radiative transfer solutions keep small grains away from the immediate

795: surface of optically thick disks and preserve temperature inversion.

796: This is achieved under the boundary condition of maximum possible

797: temperature reached by all dust grains. Then the surface becomes too

798: hot for small grains to survive, which leaves it populated only by big

799: grains. By choosing this favorable boundary condition, we made an

800: important presumption about the dust dynamics. Dust has to be

801: dynamically transported to the distances where all grains start to

802: sublimate.  This can be naturally occurring in optically thick disks

803: due to dust and gas accretion. However, in \S\ref{transverse} we

804: discovered that big grains can survive

805: closer to the star than the inner edge of optically thick disk. The

806: only requirement is that the disk becomes vertically optically thin at

807: these close distances.

808: 

809: Therefore, the dust sublimation zone is not a simple sharp step-like

810: transition, but rather a large zone (figure \ref{Disksketch}). We

811: estimate its size by considering the inner radius of optically

812: thick and thin disks \citep{VIJE06}

813: \eq{\label{Eq-Rin}

814:  R_{in} =  {\Psi R_*\over 2}\left({T_*\over T_{sub}}\right)^2

815:    = 0.0344 \Psi \left({1500K\over T_{sub}}\right)^2

816:       \sqrt{{L_*\over L_\odot}} \left[\hbox{AU}\right]

817:  }

818: where $T_*$, $R_*$ and $L_*$ are the stellar temperature, radius

819: and luminosity, respectively, and $\Psi$ is the the correction for

820: diffuse heating from the disk edge interior. Optically thick disks

821: have $\Psi=2$, while optically thin disks have $\Psi\sim 1.2$. The

822: sublimation zone exists between these two extremes, which

823: translates to $\sim 0.2$AU for Herbig Ae stars ($\sim 50L_\odot$)

824: or $\sim 0.03$AU in T Tau stars ($\sim L_\odot$). This is a

825: considerable size detectable by interferometric imaging and has to

826: be addressed in future studies. Also, its gas is enriched by

827: metals coming from the sublimated dust, which makes this zone a

828: perfect place for dust growth. Notice that, in addition to

829: sublimation, grain growth is another way of making the disk

830: optically thinner. Hence, the evolutionary role of such a large

831: sublimation zone has to be studied further in more detail.

832: 

833: Another major problem with the dynamics of big grains is that they

834: tend to settle toward the midplane, which suppresses vertical

835: expansion of the inner disk edge and leads to the failure of the disk

836: puffing model \citep{DD04,VIJE06}. On the other hand, the concurrent

837: model incorporating a dusty outflow lacks a convincing physical

838: process responsible for the formation of a dusty wind. The sublimation

839: zone may play an important role in dusty wind processes, since its small

840: optical depth and small distance from the star should also result in

841: gas properties that are more susceptible to non-gravitational forces

842: capable of launching a dusty wind (such as magnetic fields) than the

843: rest of the dusty disk.

844: 

845: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

846: \begin{figure}

847:  \epsscale{1.2}

848:  \plotone{Figure10.ps}

849:  \caption{\label{Disksketch}

850: Dust dynamics in protoplanetary disks brings dust closer to the

851: star than the inner edge of optically thick dusty disk. According

852: to the radiative transfer solution in \S\ref{transverse}, such a

853: disk structure is possible if this zone is populated only by big

854: grains and it is vertically optically thin. These conditions

855: enable disk thermal cooling and dust survival (see also figure

856: \ref{Sketch3}). The resulting sublimation zone spans from the

857: optically thin to the optically thick inner disk radius (equation

858: \ref{Eq-Rin}). Its physical size is $\sim 0.2$AU in Herbig Ae stars

859: and $\sim 0.03$AU in T Tau stars.}

860: \end{figure}

861: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

862: 

863: 

864: \section{Conclusion}

865: \label{conclusion}

866: 

867: We analyzed the temperature structure of externally heated optically

868: thick dust clouds. We focused on the recently discovered effect of

869: temperature inversion within the optically thin surface of a cloud

870: populated by big ($\ga 1\mu$m) dust grains \citep{Isella, VIJE06}. The

871: effect is produced by reprocessing the external radiation and not by

872: any additional energy source.  The inversion manifests itself as a

873: temperature increase with optical depth, before it starts to decrease

874: once the external directional radiation is completely transformed into

875: the diffuse thermal flux. Since small grains do not show this effect,

876: the open question was whether small grains would manage to suppress

877: this effect when mixed with big grains.

878: 

879: We show analytically that small grains remove the temperature

880: inversion of big grains if the overall opacity is dominated by small

881: grains. However, this does not happen in situations where the cloud is

882: close enough to the external energy source for dust to start

883: sublimating. Small grains acquire a higher temperature than bigger

884: grains and sublimate away from the immediate cloud surface.  The exact

885: grain size composition of the surface depends on the amount of

886: external flux because different grain sizes sublimate at different

887: distances from the surface. These distances are smaller than in

888: optically thin clouds because bigger grains provide shielding to

889: smaller grains from direct external radiation. We show that the

890: temperature inversion is always preserved if small grains are removed

891: from only $\tau_V\sim 1$ of the immediate surface.

892: 

893: If the boundary condition requires all dust sizes to maximize

894: their temperature, then all small grains are removed from the

895: surface layer. They can exist only within the cloud interior

896: $exp(-\tau_V)\ll 1$ where the external radiation is completely

897: absorbed. Such a condition is expected in protoplanetary disks

898: where dust accretion moves dust toward the star. The inner disk

899: radius is then defined by the largest grains, no matter what the

900: overall grain size composition, because the largest grains survive

901: the closest to the star and dictate the surface radiative

902: transfer.

903: 

904: A new problem arises in that case. Since the temperature inversion

905: keeps the very surface of the cloud below the sublimation temperature,

906: its dust can move even closer to the star. We show that this creates

907: an optically thin dusty zone inside the dust destruction radius of an

908: optically thick disk (figure \ref{Disksketch}). Only big grains can

909: survive in this zone. We estimate that its size is large enough to be

910: detected by near IR interferometry. It consists of big grains and gas

911: enriched by metals from sublimated dust, hence favorable for grain

912: growth. This shows that the geometry and structure of inner disks

913: cannot be determined by simple ad hoc boundary conditions. It requires

914: self-consistent calculations of dust dynamics combined with radiative

915: transfer calculations and dust sublimation.

916: 

917: \acknowledgments

918: 

919: Supports by the NSF grant PHY-0503584 and the W.M. Keck Foundation are

920: gratefully acknowledged.

921: 

922: 

923: \begin{thebibliography}{}

924: 

925: \bibitem[Calvet et al.(1991)]{Calvet} Calvet, N., Patino, A., Magris,

926: G.~C., \& D'Alessio, P.\ 1991, \apj, 380, 617

927: 

928: \bibitem[Carciofi et al.(2004)]{Carciofi} Carciofi, A.~C.,

929: Bjorkman, J.~E., \& Magalh{\~ a}es, A.~M.\ 2004, \apj, 604, 238

930: 

931: \bibitem[Dullemond(2002)]{D02} Dullemond, C.~P.\ 2002, 

932: \aap, 395, 853 

933: 

934: \bibitem[Dullemond et al.(2001)]{DDN} Dullemond, C.~P.,

935: Dominik, C., \& Natta, A.\ 2001, \apj, 560, 957

936: 

937: \bibitem[Dullemond \& Dominik(2004)]{DD04}

938: Dullemond, C.~P., \& Dominik, C.\ 2004, \aap, 417, 159

939: 

940: \bibitem[Efstathiou \& Rowan-Robinson(1994)]{ERR94}

941: Efstathiou, A., \& Rowan-Robinson, M.\ 1994, \mnras, 266, 212

942: 

943: \bibitem[Isella \& Natta(2005)]{Isella} Isella, A., \& Natta, A.\

944: 2005, \aap, 438, 899

945: 

946: \bibitem[Malbet \& Bertout(1991)]{Malbet} Malbet, F., \& Bertout, C.\

947: 1991, \apj, 383, 814

948: 

949: \bibitem[Mihalas(1978)]{Mihalas} Mihalas, D.\ ``Stellar atmospheres'', San

950: Francisco, W.~H.~Freeman and Co., 1978.

951: 

952: \bibitem[Monnier \& Millan-Gabet(2002)]{interfer}

953: Monnier,J.\ D. \& Millan-Gabet, R.\ 2002, \apj 579, 694

954: 

955: \bibitem[Vinkovi{\'c} et al.(2006)]{VIJE06} Vinkovi{\'c}, D.,

956: Ivezi{\'c}, {\v Z}., Jurki{\'c}, T., \& Elitzur, M.\ 2006, \apj, 636,

957: 348

958: 

959: \bibitem[Wolf(2003)]{Wolf} Wolf, S.\ 2003, \apj, 582, 859

960: 

961: 

962: \end{thebibliography}

963: 

964: \newpage

965: 

966: 

967: 

968: 

969: 

970: 

971: 

972: 

973: \end{document}

974: 

975: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

976: