1: %\documentclass[12pt,preprint]{aastex}
2: %\usepackage{emulateapj5}
3: %\citestyle{aa}
4: \documentclass[apj]{emulateapj}
5: \usepackage{apjfonts}
6:
7: \begin{document}
8: \title{Migration and growth of giant planets in self-gravitating disks with varied thermodynamics}
9: \author{Laure Fouchet$^{1}$ \& Lucio Mayer$^{2,1}$}
10: %\affil{Department of Physics, Institut f\"ur Astronomie, ETH Z\"urich, Schaffmattrasse 10, CH-8093 Z\"urich, Switzerland}
11: %\email{fouchet@phys.ethz.ch}
12: \footnotetext[1]{Department of Physics, Institut f\"ur Astronomie, ETH Z\"urich, Schaffmattrasse 10, CH-8093 Z\"urich, Switzerland, fouchet@phys.ethz.ch}
13: \footnotetext[2]{Institut f\"ur Theoretische Physik, University of Z\"urich, Winterthurerstrasse 190, CH-8057 Z\"urich, Switzerland}
14:
15: \begin{abstract}
16:
17: We report on the results of novel global high-resolution three-dimensional simulations of disk-planet interaction
18: which incorporate simultaneously realistic radiation physics and the self-gravity of the gas, as well as allowing the planet to move.
19: %Gas surrounding the planet is rapidly heated above the initial temperature as
20: %a result of shocks along the spiral arms. %triggered by its motion at distances of a few tenths of AU from the planet, and owing to compressional heating within the accretion flow at distances comparable or smaller than the Hill radius of the planet.
21: %Maximum temperatures within the Hill radius increase as the spatial resolution is increased and the flow closer to the bottom of the potential well created by the planet is better resolved. The disk is mildly viscous ($\alpha \sim 10^{-2}$) and a clear gap appears only after 100 orbits and in general some coupling between the gas flow in the disk and the planet always exists.
22: We find that thermodynamics and radiative physics have a remarkable effect on both migration and accretion of Jupiter mass planets.
23: In simulations with radiative transfer adopting flux-limited diffusion, inward migration can be
24: decreased by about $30\%$ relative to the isothermal case, while in adiabatic runs migration
25: nearly shuts off after a few tens of orbits. Migration varies
26: because the relative strength of the inner and outer spiral perturbations is affected by thermodynamics, thus changing
27: the net torque acting on the planet.
28: %If cooling is completely inhibited, such as in the adiabatic runs, migration slows down by more than an order of magnitude because the spiral arms triggered by the planet are almost completely smeared out as a result of shock heating. Similar strong effects of thermodynamics on planet migration have been recently found for the case of low mass planets.
29: Mass accretion rates on the planet can be reduced by
30: %a factor 2 to 10
31: more than an order of magnitude going from isothermal to radiative transfer and adiabatic simulations.
32: %Although we can only follow the planet for several tens of orbits in the case of radiative transfer
33: %runs, the mass accretion rate in the late stages is still high enough to allow the capture of a gaseous
34: %envelope of several Jupiter mass if extrapolated over a timescale of about $10^4$ years.
35: A circumplanetary disk always forms except in adiabatic runs. With radiative transfer the disk is sub-keplerian
36: ($v_{rot}/v_{kep} \sim 0.7$) owing to significant pressure support. We
37: discuss the effect of circumplanetary disk structure on the drift of embedded dust grains
38: and planetesimals and thus on the formation of the rocky satellites of giant planets.
39: \end{abstract}
40:
41: \keywords{Hydrodynamics - planetary systems: formation - planetary
42: systems: protoplanetary disks}
43:
44: \section{Introduction}
45:
46: In the conventional picture, giant planets form via a two-stage process, known as core-accretion. In such a model, first a massive rocky
47: core is assembled via gravitational accumulation of planetesimals in the protoplanetary disk and then the core begins to accrete
48: the surrounding gas once it has grown massive enough \citep{Pollack96,Ida04,Alibert05}.
49: The growing planet excites density waves as it moves through the disk. Such waves exert a torque
50: on the planet \citep{Lin93} whose net effect
51: is to extract angular momentum from its orbit for standard protosolar nebula models \citep{Ward97}.
52: %First a solid core forms through the gravitational accumulation of planetesimals. Second, when this core reaches a mass of about 10 Earth masses it, the core initiates runaway accretion of the surrounding gas, and reaches a mass comparable to Jupiter in a few million years. %Semi-analytical models that include a self-consistent treatment of the growing giant planet and the mass flow from the surrounding protoplanetary disk via viscous processes are able to form planets with masses comparable to extrasolar planets before the gas disk is dissipated \citep{Alibert05}.
53:
54: %Semi-analytical models \citet{Alibert05} reproduce reasonably well the properties of observed extrasolar planets
55: %provided that
56: %The properties of the planets are reasonably consistent
57: %with observed extrasolar planets. However, these models
58: %need to postulate that disk-planet interaction does
59: %not lead to the fast inward migration that is typically observed in numerical simulations (Ward, ..Kley..., Bate).
60: %Without such assumption, planets would sink towards the central star on a characteristic timescale much shorter
61: %than their accretion timescale. %, being unable to acquire the mass of a giant planet before then.
62:
63: Different regimes of migration have been identified by numerous studies depending on the planet's mass and on the
64: importance of co-rotation torques \citep[see][for a review]{PPV}. In particular, when the planet has a mass comparable to Saturn or
65: larger the interaction with the disk becomes markedly nonlinear %. In
66: %the conventional
67: (type II migration) %theory such a massive
68: and the planet
69: carves a gap in the disk as a result of the planetary tide \citep{Lin93,Crida07}.
70: %After the gap forms
71: Then, the planet decouples dynamically from the disk %, cannot
72: %feel anymore the back reaction of density waves but rather
73: and migrates inward at a pace determined by the local viscous timescale
74: \citep{Lin86}.
75: %A third regime, type III regime, which is due to
76: %corotation torques produced by the gas closest to the planet, has been identified more recently
77: %\citep{Masset03} {\bf (Masset, Crida \& Morbidelli)??}. This leads to the fastest migration rate.
78: %Corotation torques are caused by the interaction with gas contained within the Hill radius of the planet itself. Migration in this latter regime can be inward or outward depending on the details of the protoplanetary disk profile; if it is inward it can bring the planet to the center in less than $10^4$ years.
79: %Despite the vigorous effort done in modeling disk-planet interaction, the current understanding of migration in both the linear
80: %and nonlinear regime is mostly based on calculations that neglect several physical ingredients (Papaloizou et al. PPV).
81:
82: %For example, \citet{Nelson05} has shown that the inclusion of magnetic fields can produce stochastic migration of small planets, with the result that the net inward migration rate of low mass planets is appreciably reduced.
83: A major assumption of almost all simulations published so far is that the disk is locally isothermal. This means that any heating
84: is immediately radiated away, which would only be true if the disk was optically thin.
85: %A steady-state balance between heating and cooling, implicit in the isothermal assumption, is not correct in the midplane of the standard protoplanetary disk model adopted in the literature, a minimum solar nebula model with mass $\sim 0.01 M_{\odot}$.
86: The planet moves supersonically in the disk, shock-heating the gas, and gas accretion onto the planet generates compressional heating.
87: With typical densities $10^{-11}$g/cm$^3$, the midplane of a minimum solar nebula disk is indeed optically thick ($\tau \approx 10$)
88: \citep{d'Angelo03}.
89: Recently, Paardekooper \& Mellema (2006, 2008, hereafter \citet{Paardekooper06,Paardekooper08}) have used both 2D and 3D adaptive mesh refinement simulations with radiative
90: transfer modeled via flux-limited diffusion finding that for low mass planets both inward migration and gas accretion can be strongly
91: suppressed.
92: %both migration and gas accretion for low mass planets in the Type I regime are significantly affected by the inclusion of radiative transfer.
93: Klahr \& Kley (2006, hereafter \citet{KK06}) studied
94: Jupiter-sized planets with a static 3D grid code and flux-limited diffusion. They did not report
95: significant differences on migration compared with the isothermal case but found the structural
96: evolution of the circumplanetary gas distribution to be strongly affected by radiation physics as
97: noticed earlier in the 2D nested-grid calculations by \citet{d'Angelo03}.
98:
99: %Indeed they find that no circumplanetary disk forms
100: %with radiative transfer, while this is seen in isothermal simulations. Whether a disk is present or not
101: %has implications on the gas accretion onto the planet and on the eventual formation of satellites of giant.
102:
103: Yet, even these recent simulations lack several important ingredients. First, they do not treat self-consistently the dynamics of the disk,
104: planet and star; the planet and the star cannot move, and in some cases a gap is introduced already at the beginning of the simulation (KK06).
105: Second, except for the 2D simulations of d'Angelo et al. (2003), they adopt inviscid
106: disks. Recently, Edgar (2007) found that giant planet migration in an isothermal viscous disk does not obey the standard type II
107: regime. A deep gap is never produced and hence migration does not proceed on the viscous timescale.
108: Finally, all these simulations neglect the self-gravity of the gas, which is required when simulating freely
109: moving planets and might affect disk torques (Baruteau \& Masset 2008). Self-gravity has been previously included only in
110: a few isothermal calculations \citep{Nelson03a,Nelson03b,Lufkin04}.
111:
112: In this Letter we present the first high-resolution three-dimensional hydrodynamical simulations of the interaction between a massive,
113: Jupiter-sized planet and a surrounding viscous protoplanetary disk that include simultaneously radiative transfer, shock heating,
114: self-gravity of the gas and fully self-consistent dynamics. We study migration, mass flow towards the planet
115: and the circumplanetary gas distribution, exploring also the implications on the formation of satellites of giant planets.
116: We compare the results with those obtained for locally isothermal disks as well as other simulations with simplified
117: disk thermodynamics.
118:
119: \begin{table}
120: \scriptsize
121: \caption{Simulations parameters\label{tab:M}}
122: \begin{tabular}{rrrrrr}
123: \tableline
124: Run & Number & Mean & Partial & Mean & Partial \\
125: Name & of & accretion & accretion & migration & migration \\
126: & orbits & rate & rate & rate & rate \\
127: \tableline
128: Model M1 & M$_d$= & 0.004 M$_{Sol}$ & & & \\
129: \tableline
130: %IsoT1M1 & 39.8 & 6.05 10$^{-7}$ & 7.25 10$^{-7}$ & 1.26 10$^{-3}$ & 5.02 10$^{-4}$ \\
131: %Adia1M1 & 80.4 & 9.78 10$^{-10}$ & 4.89 10$^{-9}$ & 7.29 10$^{-5}$ & 3.24 10$^{-4}$ \\
132: %%FLD1M1 & 0.7 & \nodata & \nodata & \nodata & \nodata \\
133: %$\star$ IsoT200K1 & 173.4 & 5.28 10$^{-8}$ & 2.35 10$^{-7}$ & 2.87 10$^{-4}$ & 6.09 10$^{-4}$ \\
134: %$\star$ FLD200K1 & 20.6 & 2.09 10$^{-8}$ & 2.09 10$^{-8}$ & 4.22 10$^{-4}$ & 4.22 10$^{-4}$ \\
135: %$\star$ NIsoT200K1 & 82.3 & 0 & 0 & 6.07 10$^{-5}$ & 3.16 10$^{-4}$ \\
136: %$\star$ Adia200K1 & 93.2 & 0 & 0 & 3.73 10$^{-5}$ & 1.83 10$^{-4}$ \\
137: %\tableline
138: %Model M2 & M$_d$= & 0.01 M$_{Sol}$ & & & \\
139: %\tableline
140: %IsoT1M2 & 38.9 & 1.56 10$^{-6}$ & 1.60 10$^{-6}$ & 1.34 10$^{-3}$ & 4.01 10$^{-4}$ \\
141: %FLD1M2 & 2.9 & 9.33 10$^{-7}$ & \nodata & 1.15 10$^{-3}$ & \nodata \\
142: %NIsoT1M2 & 14.3 & 2.88 10$^{-7}$ & \nodata & 1.22 10$^{-3}$ & \nodata \\
143: %Adia1M2 & 17.4 & 2.19 10$^{-8}$ & \nodata & 9.01 10$^{-4}$ & \nodata \\
144: %IsoT100K2 & 167.7 & 1.21 10$^{-6}$ & 1.82 10$^{-6}$ & 6.48 10$^{-4}$ & 9.60 10$^{-4}$ \\
145: %FLD100K2 & 27.9 & 1.09 10$^{-6}$ & 1.09 10$^{-6}$ & 8.33 10$^{-4}$ & 8.33 10$^{-4}$ \\
146: %NIsoT100K2 & 111.8 & 0 & 6.46 10$^{-10}$ & 9.61 10$^{-5}$ & 4.00 10$^{-4}$ \\
147: %Adia100K2 & 133.8 & 0 & 0 & 1.43 10$^{-4}$ & 2.24 10$^{-4}$ \\
148: %
149: IsoT1M1 & 39.8 & 6.34 10$^{-4}$ & 7.59 10$^{-4}$ & 1.26 10$^{-3}$ & 5.02 10$^{-4}$ \\
150: Adia1M1 & 80.4 & 1.02 10$^{-6}$ & 5.12 10$^{-6}$ & 7.29 10$^{-5}$ & 3.24 10$^{-4}$ \\
151: %FLD1M1 & 0.7 & \nodata & \nodata & \nodata & \nodata \\
152: $\star$ IsoT200K1 & 173.4 & 5.53 10$^{-5}$ & 2.46 10$^{-4}$ & 2.87 10$^{-4}$ & 6.09 10$^{-4}$ \\
153: $\star$ FLD200K1 & 20.6 & 2.19 10$^{-5}$ & 2.19 10$^{-5}$ & 4.22 10$^{-4}$ & 4.22 10$^{-4}$ \\
154: $\star$ NIsoT200K1 & 82.3 & 0 & 0 & 6.07 10$^{-5}$ & 3.16 10$^{-4}$ \\
155: $\star$ Adia200K1 & 93.2 & 0 & 0 & 3.73 10$^{-5}$ & 1.83 10$^{-4}$ \\
156: \tableline
157: Model M2 & M$_d$= & 0.01 M$_{Sol}$ & & & \\
158: \tableline
159: IsoT1M2 & 38.9 & 1.63 10$^{-3}$ & 1.68 10$^{-3}$ & 1.34 10$^{-3}$ & 4.01 10$^{-4}$ \\
160: FLD1M2 & 2.9 & 9.77 10$^{-4}$ & \nodata & 1.15 10$^{-3}$ & \nodata \\
161: NIsoT1M2 & 14.3 & 3.02 10$^{-4}$ & \nodata & 1.22 10$^{-3}$ & \nodata \\
162: Adia1M2 & 17.4 & 2.29 10$^{-5}$ & \nodata & 9.01 10$^{-4}$ & \nodata \\
163: IsoT100K2 & 167.7 & 1.27 10$^{-3}$ & 1.91 10$^{-3}$ & 6.48 10$^{-4}$ & 9.60 10$^{-4}$ \\
164: FLD100K2 & 27.9 & 1.14 10$^{-3}$ & 1.14 10$^{-3}$ & 8.33 10$^{-4}$ & 8.33 10$^{-4}$ \\
165: NIsoT100K2 & 111.8 & 0 & 6.77 10$^{-7}$ & 9.61 10$^{-5}$ & 4.00 10$^{-4}$ \\
166: Adia100K2 & 133.8 & 0 & 0 & 1.43 10$^{-4}$ & 2.24 10$^{-4}$ \\
167: \tableline
168: \end{tabular}
169: \tablenotemark{k}
170: \tablenotetext{k}{The Table shows the accretion and migration rates for the various
171: runs. The legend for the run names is as follows: IsoT refers to
172: Isothermal runs, FLD to runs with the Flux
173: Limited Diffusion approximation, NIsoT are adiabatic runs with
174: $\gamma$ = 6/5 and Adia are adiabatic runs with $\gamma$ = 7/5.
175: 100K corresponds to 100 000 particles, 200K to 200 000
176: particles and 1M to 1 Million particles. The runs are normally done
177: with a planet gravitationnal softening of R$_H$/5, where
178: R$_H$=0.35~AU is the Hill radius for a Jovian planet, except for
179: those marked with $\star$ for which the softening is equal to R$_H$. Accretion rates are given in M$_{J}$ yr$^{-1}$ and migration
180: rates in AU yr$^{-1}$. Mean rates are computed by
181: averaging over time over the full extent of the simulation. Partial rates are computed after a number of
182: orbits equivalent to the maximum
183: number of orbits for the FLD runs, namely after 20.6 orbits for
184: model M1 and 27.9 orbits for model M2.
185: %Rates are computed by taking the final position/mass minus the
186: %initial ones and dividing by the elapsed time.
187: When the accretion rate is 0, it means that the mass
188: inside the Hill radius is below the SPH mass resolution.}
189: \end{table}
190:
191: %\medskip
192: %{\centering
193: %\epsfxsize=4.2truecm
194: %\epsfbox{Images/TransRad1M_00045_Md0004_rhomid_xy.eps}
195: %\epsfxsize=4.2truecm
196: %\epsfbox{Images/TransRad1M_00045_Md0004_temp_xy.eps}
197: %\figcaption[angle45.ps]{\label{fig:angle45}
198: %Density (top) and temperature maps (bottom)
199: %of a flux-limited diffusion run after 0.7 orbits. The shock along the spiral arms
200: %triggered by the planet is evident. The planet is seen as a hot spot in the disk.}}
201: %\medskip
202:
203: \begin{figure}%[h]
204: %\plottwo{Images/rhomid_xy_Md0004_FLD1M_00045_pl_07.eps}{Images/temp_xy_stdcol_Md0004_FLD1M_00045_pl_07.eps} % Md0004 => 0.7 orbits
205: %\plottwo{Images/rhomid_xy_Md001_FLD1M_00060_pl_07.eps}{Images/temp_xy_stdcol_Md001_FLD1M_00060_pl_07.eps}
206: %\includegraphics[width=4.2cm]{Images/rhomid_xy_Md001_FLD1M_00060_pl_07.eps}
207: %\plottwo{Images/GapProf2_Md001_LR_00585.eps}{Images/temp_xy_stdcol_Md001_FLD1M_00060_pl_07.eps}
208: %\includegraphics[width=4cm]{Images/GapProf2_Md001_LR_00585.eps}
209: %\includegraphics[width=4.1cm]{Images/temp_xy_stdcol_Md001_FLD1M_00060_pl_07.eps}
210: %\includegraphics[width=4cm]{f1a.eps}
211: %\includegraphics[width=4.1cm]{f1b.eps}
212: %\plotone{f1a.eps}
213: \plotone{f1.eps}
214: %\caption{Left: Gap profile for runs IsoT100K2, FLD100K2,
215: % NIsoT100K2 and Adia100K2 after 27.9 orbits and IsoT100K2 also after
216: % 190 orbits. The legend for the different curves is in the figure. Right: Temperature map with overplotted midplane
217: % density contours of run FLD1M2 after 2.9 orbits. The shock along the spiral arms triggered by the planet is evident. The planet is seen as a hot spot in% the disk.}
218: \caption{Temperature map with overplotted midplane
219: density contours of run FLD1M2 after 2.9 orbits. The shock along the spiral arms triggered by the planet is evident. The planet is seen as a hot spot in the disk.}
220: \label{fig:ArmFLD}
221: \end{figure}
222:
223: \section{Initial Conditions and Simulations}
224:
225: We have run different simulations with varying disk masses, resolution and thermodynamics (see Table~1).
226: Our simulations include the self-gravity of the gas \citep{Nelson03a,Nelson03b,Lufkin04}.
227: We place a 1 Jupiter mass ($1 M_J$) planet on a circular orbit at $5$~AU for
228: model M1 (resp. $5.4$~AU for model M2) from a
229: $1 M_{\odot}$ mass star in an initially axisymmetric disk.
230: The planet and the star are represented by softened N-Body particles free to move in response
231: to their mutual gravity and that of the disk. There is no gap initially.
232: Mass is allowed to accumulate around the planet since we do not use a prescription for accretion.
233: Our initial disk extends from %0.4 to 4 in code units, which is
234: 2 to 20~AU in model M1 and from 1 to 25~AU in model M2.
235: %As the disc viscously spreads, the particles quickly fill the central hole because no accretion on the star is included. After 14 orbits, the central hole
236: %is already 0.03 in code units, i.e. 0.15 AU.\\
237: %We adopt initial surface density of $\Sigma(5~AU)=75~\mathrm{g.cm^{-2}}$.
238: The surface density profile is $\Sigma = \Sigma(5~AU)
239: (r/5~AU)^{-1.5}$. We adopt an initial surface density of
240: %either $\Sigma(5~AU)=75~\mathrm{g.cm^{-2}}$ or $\Sigma(5~AU)=XXX~\mathrm{g.cm^{-2}}$, which yield a total disk mass of $M_\mathrm{disk} = 0.004~M_{\sun}$ (model M1) or $M_\mathrm{disk} = 0.01~M_{\sun}$ (model M2).
241: $\Sigma(5~AU)=75~\mathrm{g\ cm^{-2}}$ (resp. $150~\mathrm{g\ cm^{-2}}$),
242: which yield a total disk mass of $M_\mathrm{disk} = 0.004~M_{\sun}$ for model M1
243: and 0.01~$M_{\sun}$ for model M2.
244: The vertical density profile is initially gaussian.
245: We assume the same disk temperature profile as in \citet{KK06} for the sake of comparison, $T=T_0 r^{-1}$ ($T$ is independent on z).
246: The pressure scale-height is initially chosen to be $H/r = 0.05$ and implies $T{_0} = 102$ K.
247: %We use $10^5$, $2 \times 10^5$ or $10^6$ gas particles in our runs.
248: %The softening of the planet varies between $1$ and $1/5$ of the Hill radius in different simulations, with the
249: %smallest values being adopted in the simulations with the highest mass resolution (number of particles).
250:
251: %The Toomre parameter $Q =\frac{c_\mathrm{s} \kappa}{\pi G \Sigma}$
252: %with $c_\mathrm{s}$ the sound speed, $\kappa$ the epicyclic frequency, $G$ the gravity constant
253: %and $\Sigma$ the surface density, is $ > 1.4$ everywhere in the disk, hence the disk is
254: %gravitationally stable. Self-gravity, however, might still play a role in disk-planet interaction
255: %\citep{Nelson03a,Nelson03b,Lufkin04}.
256:
257: As for disk thermodynamics, we consider the following different cases:
258: (1) locally isothermal (IsoT) runs, in which
259: the temperature of individual particles is kept constant over time;
260: (2) adiabatic (Adia) and nearly isothermal (NIsoT) runs,
261: in which the gas is evolved adiabatically assuming an adiabatic index of, respectively, $\gamma=7/5$ and $\gamma=6/5$
262: and solving the internal energy equation including shock heating; (3)
263: radiative transfer (FLD) runs, in which we solve the energy
264: equation with added flux-limited
265: diffusion and blackbody radiation at the disk edge to model the
266: thermal energy flow in the disk, as described in \citet{Mayer07}. We adopt realistic
267: opacities by d'Alessio et al. (1997).
268:
269: \begin{figure}%[h]
270: %%\includegraphics[width=4.5cm]{Images/ArmVol_x_EOS_Md001_LR_00585_all.eps}
271: %%\plottwo{Images/ae_Md001_LR.eps}{Images/Torques_Md001_LR_00585.eps}
272: \plottwo{f2a.eps}{f2b.eps}
273: %\plotone{f2a.eps}
274: %\plotone{f2b.eps}
275: %%\includegraphics[width=4.5cm]{Images/ae_Md001_LR.eps}
276: %%\hspace{0.2cm}
277: %%\includegraphics[width=3.6cm]{Images/Torques_Md001_LR_00585.eps}
278: %\caption{Left: Amplitude of the gas density in the spiral arm for FLD
279: % and isothermal (left axis scale) and ratio of FLD versus isothermal
280: % (dotted line, right axis scale). The negative x correpond to the
281: % outer arm and the positive ones to the inner, denser arm. Right:
282: % Torques exerted by the disk on the planet, showing the effect of
283: % thermodynamics on disk-planet interaction at large scales. (Model
284: % M2, low resolution, after 27.9 orbits)}
285: \caption{Semi-major axis of the planet versus time (left) and
286: azimuthally averaged radial torques exerted by the disk on the
287: planet (right) as a function of the distance from the star for runs IsoT100K2, FLD100K2,
288: NIsoT100K2 and Adia100K2 after 20 orbits. Note that, in the adiabatic case, the inner and outer torques nearly balance. The legend for the different curves is in the figure. The position of the
289: planet is shown with symbols of different colors depending on the
290: run (right panel).}
291: \label{fig:MigrTorque}
292: \end{figure}
293:
294: The simulations are performed with the parallel 3D Tree+SPH code GASOLINE \citep{Wadsley04},
295: which includes adaptive multiple timesteps. In radiative transfer runs radiative timesteps
296: can become very short so that only a few tens of orbits could be
297: explored \citep[see also][]{Paardekooper06,Paardekooper08}.
298: %This is however a larger
299: %number of orbits compared to previous work using spatially adaptive codes
300: %with similarly sophisticated radiation physics \citep{Paardekooper06,Paardekooper08}.
301: We use the standard Monaghan viscosity with coefficients $\alpha=1$ and $\beta=2$ and with the Balsara switch
302: to reduce viscosity significantly in shear flows \citep{Balsara95}. We find an average equivalent
303: effective Shakura-Sunyaev $\alpha_\mathrm{SS}$ parameter
304: $\sim 10^{-2}$, although this varies appreciably with time and location in the disk.
305: For comparison, protoplanetary disks are expected to have $\alpha_\mathrm{SS}$ between $10^{-4}$ and $10^{-1}$ as
306: a result of MHD turbulence (Papaloizou \& Nelson 2003).\\
307: %Finally, we do not take into account accretion onto the planet as done in some of the runs by \citet{KK06},
308: %instead we let the gas accumulate around the planet (simulations with sink particles that treat accretion are
309: %currently under way).
310:
311: \section{Results}
312:
313: We find that both the orbital evolution and the properties of the gas flow around the planet depend significantly on disk thermodynamics.
314: %Another
315: %parameter of the simulations that has a non-negligible influence is the gravitational softening of the planet, as previously
316: %indicated in the literature (...). The latter is however most crucial for the less realistic among the runs considered in this
317: %paper, namley the isothermal runs. In a forthcoming paper we will discuss in detail the effect of gravitational softening and
318: %other aspects of isothermal, self-gravitating disks. Here we focus on comparing runs with identical mass resolution and
319: %gravitational softening of the planet but different thermodynamics.
320:
321: \subsection{Migration}
322:
323:
324: Our reference case for the study of migration is represented by the locally isothermal runs. These have been widely used in the literature,
325: although only rarely in the past self-gravity was included and/or the planet was allowed to move \citep{PPV,Bate03,Nelson03a,Nelson03b}.
326: The planet carves a gap quite slowly compared to simulations of inviscid disks (e.g. Nelson et al. 2003a).
327: However, the drop in the surface density near the planet
328: reaches more than order order of magnitude after 190 orbits, in agreement with the results of
329: other three-dimensional grid-based
330: and SPH simulations of Jupiter-mass planets in disk with moderate viscosity \citep{deValBorro06}.
331: At this late stage the migration timescale is $\sim 3 \times 10^4$ years, somewhat faster than the timescale
332: of $\sim 8 \times 10^4$ years typically inferred for Jupiter mass planets evolving in inviscid disks
333: on a fixed orbit \citep[e.g.][]{PPV,KK06,Edgar07}. A speed-up of the migration rate
334: by a factor $2-3$ for disks with $\alpha$ viscosities $\sim 10^{-2}$
335: has already been reported in the literature (d'Angelo et al. 2003;
336: Edgar 2007). For the first few tens of orbits, when the gap is barely opened, the planet excites strong
337: spiral modes since it is still well coupled with the disk and the migration rate is
338: much faster than in the late stage (see partial migration rate in Table~1). Self-gravity also should lead to an
339: acceleration of migration before the planet has opened a gap
340: (Baruteau \& Masset 2008).
341: The mean migration rate increases with increasing disk mass (Table~1), in agreement with Edgar (2007), who also studied viscous isothermal
342: disks that do not open deep gaps.
343:
344: In flux-limited diffusion (FLD) runs the planet has a significant effect on the temperature structure of the disk
345: around it. The triggered spiral arms are associated with a shock since the planet is moving supersonically. The
346: temperature along the spiral arms thus rises significantly because the disk midplane is optically thick (Figure~1) an effect which
347: is obviously absent in isothermal runs.
348: Flux-limited diffusion runs are followed up to a point when they are
349: still in the weak gap regime, i.e. for no more
350: than 30 orbits. The gap is somewhat weaker in these runs because the disk is heated by the spiral shocks in the region around the planet, becoming thicker and thus making it harder to satisfy the gap opening condition.
351:
352: %In such a configuration the planet is not expected to behave as in the classic type II regime.
353: %however, after an equivalent number of orbits the depth of the gap in the flux-limited
354: %diffusion and isothermal runs appears almost identical. The similarity of the gaps is due to the fact that the temperature
355: %of the gas near the planet, and therefore the local thickness of the disk, is comparable between the two type of runs,
356: %which implies a similar configuration relative to the standard gap-opening condition {\bf (see section 3.1)??}.
357: %After a few tens of orbits migration can slow down or accelerate depending on how well resolved is the gas flow around the
358: %planet. It typically accelerates in the highest resolution runs (higher number of particles and smaller softening) in which the
359: %circumplanetary disk is well resolved. This behaviour is conistent with the findings of ... regarding Type III migration. The
360: %presence of self-gravity likely exacerbates the effect of torques from nearby gas (Masset recent...).
361:
362: Despite the shallower gap, in FLD runs migration is slower than in the isothermal case, by $30\%$ for the case of the low
363: disk mass, and by nearly $15\%$ for the case of the low mass disk (Figure~2, left panel and Table~1).
364: The slower migration is due to an effect already seen and
365: explained in \citet{Paardekooper08} for deeply embedded, low-mass planets. It is due to the change of the balance between the outer
366: and inner torque acting on the planet as the thermodynamics is varied.
367: %This change of torque balance is in turn driven by a variation of the strength of the outer and inner spiral modes triggered by the planet.
368: In particular, we find that, the net torque is more negative in the
369: isothermal runs relative to the FLD ones
370: %while the outer torque varies little between the isothermal and the FLD runs, the inner torque, which tends
371: %to push the planet outward, is appreciably stronger in the FLD runs
372: (Figure~\ref{fig:MigrTorque}, right panel). The variation of torque balance is due to a variation
373: in the structure of the spiral perturbations excited by the planet.
374: %The difference between isothermal and radiative transfer runs is quantitatively less than seen by \citet{Paardekooper08} for deeply embedded low-mass planets.
375:
376: Torques are even more affected by the different temperature and density evolution in the adiabatic runs (Figure~2, left panel). In these runs the gas
377: along the spiral shocks rises its temperature by up to a factor of 5 and adiabatically expands away from the midplane.
378: As a result, after 50 orbits the gas surface density in the disk midplane around
379: the planet has decreased by a factor $12.5$ in the run with $\gamma =7/5$ and
380: by $8.3$ in the run with $\gamma = 6/5$. Because of the decreased gas surface density, the spiral arms around the planet weaken
381: considerably with time, and so do the torques exerted by them. Then, after a few tens of orbits, migration nearly shuts off (Figure~2, right panel).
382: %At this time the disk has become
383: %very thick and featureless.
384: The planet is never able to open a gap because the radius
385: of the planet's Roche lobe is smaller than the local disk semi-thickness.
386:
387: %This is confirmed by a quantitative Fourier analysis of non-axisymmetric modes in the disk. This analysis shows
388: %that the $m=2$ mode associated with the planet-triggered double-armed spiral has an amplitude of XXX for the $\gamma = 7/5$
389: %run (XXX for the $\gamma=4/3$ run) after 50 orbits compared to an amplitude $...$ for an isothermal run with identical
390: %softening and particle number. The modes become substantially weaker after about 10 orbits. The strength of the
391: %modes is shown in Figure 1.
392:
393: \begin{figure}[h]
394: %%\medskip
395: %%{\centering
396: %%\epsfxsize=9truecm
397: %%\epsfbox{circumaps.ps}
398: %%\plottwo{Images/temp_xy_stdcol_Md001_FLD1M_00060_pl.eps}{Images/temp_rz_stdcol_Md001_FLD1M_00060_pl.eps}
399: %\plottwo{f3a.eps}{f3b.eps}
400: %\plotone{f3a.eps}
401: %\plotone{f3b.eps}
402: %%\includegraphics[width=4cm]{Images/temp_xy_Md001_FLD1M_00060_pl.eps}
403: %%\includegraphics[width=4cm]{Images/temp_rz_Md001_FLD1M_00060_pl.eps}
404: %%\plottwo{Images/temp_xy_stdcol_Md001_IsoT1M_00060_pl.eps}{Images/temp_rz_stdcol_Md001_IsoT1M_00060_pl.eps}
405: %\plottwo{f3c.eps}{f3d.eps}
406: %\plotone{f3c.eps}
407: %\plotone{f3d.eps}
408: %\plotone{f3.eps}
409: \plottwo{f3b.eps}{f3d.eps}
410: %%\figcaption[angle45.ps]{\label{fig:maps}
411: %%Temperature maps of circumplanetary gas with superimposed density countours for FLD and isothermal runs
412: %%(left, xy projection; right, xz projection) (model M2, high resolution, after 2.9 orbits).}}
413: %%\medskip
414: %\caption{Temperature maps of circumplanetary gas with superimposed
415: % density countours for FLD1M2 (top) and IsoT1M2 (bottom) runs after
416: % 2.9 orbits (left, xy midplane slice; right, azimuthally averaged rz
417: % map). The colorbar gives the temperature on a logarithmic scale in K.}
418: \caption{Azimuthally averaged temperature maps of circumplanetary gas with superimposed
419: density countours for FLD1M2 (left) and IsoT1M2 (right) runs after
420: 2.9 orbits. The colorbar gives the temperature on a logarithmic scale in K.}
421: \label{fig:maps}
422: \end{figure}
423:
424: \subsection{Gas accretion and planetary growth}
425:
426: Since we do not include explicitly accretion onto the planet in our simulations we calculate an hypothetical accretion rate by
427: measuring the mass of gas that accumulates within the Hill radius of the planet ($R_H = 0.35$ AU) as a function of time.
428: %the smallest resolvable distance from the planet, which we choose to
429: %be equivalent to two planet's softening lengths (the gravitational
430: %potential is well resolved only down to two softening lengths for the
431: %spline kernel softening adopted in GASOLINE).
432: The accretion rate thus defined is found to be strongly dependent on disk thermodynamics as well as on the disk mass, and,
433: not surprisingly, increases moderately as the gravitational softening of the planet is decreased (see Table~1).
434:
435: In isothermal M1 runs the accretion rate averaged over nearly 200 orbits is $5 \times 10^{-5}$ M$_{J}$ yr$^{-1}$.
436: in agreement with previous estimates obtained for similar disk masses and effective $\alpha$ viscosities
437: (e.g. Bate et al. 2003; d'Angelo et al. 2003).
438: \citet{KK06} compute mass accretion rates about 3 times smaller for planets
439: in an isothermal disk but their disk is inviscid and they start with a gap,
440: both differences going in the direction of lowering somewhat the mass accretion.
441: %For the higher disk mass, M2, the accretion rate grows by more than an order of magnitude higher. This is
442: %due not only to the different disk mass but also to the smaller gravitational softening of the planet
443: %that yields a deeper planetary potential well. Indeed, variations of the rate are less than $30\%$ when comparing
444: %different mass resolutions (models M2) and about a factor of 3 when both the mass resolution and the gravitational
445: %softening are changed (models M1).
446:
447: As seen in \citet{Paardekooper08} for the case of low mass planets, planetary accretion in non-isothermal disks
448: can be reduced compared to isothermal disks. This is because the gas is compressionally heated as it flows towards
449: the planet, and the resulting pressure gradient opposes further gas inflow. The temperature structure around
450: the planet in one of the FLD runs is shown in Figure~\ref{fig:maps}.
451: For our lowest disk mass (M1) the accretion rate over the first 20 orbits is
452: $\sim 2 \times 10^{-5}$ M$_{J}$ yr$^{-1}$ in the FLD run, 20 times lower
453: than for the isothermal during the same part of the orbital evolution (Table~1), and comparable with what
454: \citet{KK06} find for their radiative transfer runs starting with or without a gap.
455: Similar rates were also obtained by d'Angelo et al. (2003) for their radiative viscous models.
456: The difference between isothermal and FLD runs is smaller in the case of the largest disk mass (M2).
457: In the FLD runs with in the M2 disk, which have the smallest softening, the planet should
458: be able to accrete a few Jupiter masses over $\sim 10^4$ years, a time comparable to the migration rate.
459: %However, long term calculations
460: %treating planets as accreting sink particles with proper boundary conditions for the thermal energy
461: %flow will be necessary in order to study robustly the mass growth of the planet.
462: %The temperature near the planet, and thus the pressure, increases as the resolution is increased,
463: %in agreement with \citet{KK06}.
464:
465: %The different accretion rate between isothermal and FLD runs is an effect of the different thermodynamics,
466: %In FLD runs gas is compressionally heated as it falls towards the planet and radiative cooling is not fast enough
467: %to dissipate such heating because the disk midplane is optically thick.
468: %A strong pressure gradient is thus generated around the planet that opposses further mass accumulation.
469:
470: %%%perhaps double check res. effect%%%
471:
472: %In FLD runs the temperature raise
473: %near the planet is barely affected by changing the efficiency of radiation at the disk edge.
474: %Significantly lower mass accretion rates for non-isothermal runs were recently found by \citep{Paardekooper06,Paardekooper08}
475: %for lower mass planets.
476:
477: %Lowering the efficiency of blackbody cooling in the optically thin region of the disk by
478:
479: %a factor of 5 has almost no effect on the temperature structure near the planet, and thus on the accretion rate,
480:
481: %since this region is dominated by compressional heating and encompasses gas with high optical depths ($\tau > 10$).
482:
483: In adiabatic disks accretion, like migration, essentially shuts off after a few tens of orbits (Table~1). Due to the absence of cooling
484: compressional heating in this case is indeed more effective. Moreover, heating due to spiral shocks raises
485: the temperature of the gas even at some distance from the planet, which further stifles accretion.
486: %The gas in a region of with size $> 1$ AU around the planet becomes so hot and pressure supported to form a large envelope naerly in hydrostatic equilibrium.
487:
488: %The gravity of the planet is simply not strong enough to win over gas pressure and restart and accretion flow.
489:
490: \subsection{The circumplanetary disk}
491:
492: A circumplanetary disk begins to form already during the first few orbits in both the isothermal and the radiative transfer runs (Figure~3).
493: The disk is thicker in the radiative transfer runs due to higher thermal pressure. It is sub-keplerian in both
494: cases. Yet, there is a fairly extended region, between $0.15$ and $0.35$~AU from the planet, within which
495: the gas azimuthal velocity is close to $70\%$ of the local keplerian velocity (Figure~4, left panel).
496: %gas is still quite close to keplerian motion
497: %(see Figure \ref{fig:sopt} left).
498: %The structure of the disk is shown
499: %by the density contours shown in Figure \ref{fig:maps}.
500: In the adiabatic runs the gas
501: becomes hot enough to form a diffuse pressure supported atmosphere around the planet rather than a dense circumplanetary
502: disk. In the latter case the gas is markedly sub-keplerian (Figure~4, left panel). In reality this is an extreme configuration that
503: could occur only if cooling is completely suppressed by some stronger heating mechanism. A protoplanetary disk strongly irradiated
504: by neighboring OB stars in a dense cluster might present such extreme conditions at large radii, $R > 10$ AU.
505: \citep{Alexander06}.
506:
507: \begin{figure}
508: \plottwo{f4a.eps}{f4b.eps}
509: %\plotone{f4a.eps}
510: %\plotone{f4b.eps}
511: %\plotone{f4.eps}
512: \caption{Gas azimuthal velocity divided by the keplerian velocity
513: (left) and grain size for which $T_s/T_{orb}$=1/(2 $\pi$), which
514: correpond to the fastest possible migration towards pressure maxima
515: (right) for runs IsoT100K2, FLD100K2, NIsoT100K2 and Adia100K2 after
516: 27.9 orbits. The legend for the different curves is in the right panel.}
517: \label{fig:sopt}
518: \end{figure}
519:
520: The velocity, density and temperature of the gas flow in the circumplanetary disk can determine whether rocky satellites of giant
521: planets may form or not. Based on the strong sub-keplerianity of the
522: gas around their Jupiter-mass planet, \citet{KK06}
523: concluded that satellites could not form in their radiative transfer simulations.
524:
525: In Figure \ref{fig:sopt}, right panel, we show the particle size for which $T_s/T_{orb}$=1/(2 $\pi$) where $T_s$ is the stopping time in the Epstein regime and $T_{orb}$ is local the orbital time.
526: For any location in the circumplanetary disk grains whose size obeys the latter relation are the fastest to migrate towards pressure maxima
527: (i.e. towards the planet in this case given that pressure
528: decreases monotonically from the planet outwards).
529: Much smaller particles will be entrained with the gas and migrate slowlier to the planet, while much larger boulders will follow
530: perturbed keplerian orbits around the planet \citep{Weidenschilling77}.
531: For isothermal and FLD runs the fastest migrating grains have sizes in the range 0.2 to 0.8~m, and the stopping time is
532: almost the same in both cases.
533: %We also show the maximal particle size for which the Epstein regime is still valid%
534: We note that our calculation is strictly valid only down to $0.15$ AU from the planet; at smaller distances
535: the assumption of the Epstein regime breaks down.
536: For the adiabatic case, the Epstein regime is valid for bodies up to 10~m in size because the density of the circumplanetary envelope is very low.
537: However, since the gas there is very hot, the fastest migrating grains are smaller than for the isothermal and FLD cases,
538: having sizes in the range 0.06 to 0.1~m.
539:
540: %On Figure \ref{fig:sopt} left, we see that the gas is noticeably subkeplerian in the adiabatic case, but the density is so low that it only affects smaller grains. We also note that gas closer to keplerian in the Isothermal case than in the FLD one. Still, between 0.15 and 0.35~AU from the planet, the gas azimuthal velocity is never less than 0.7 of keplerian.
541: Based on these results, we can expect that only very large boulders, i.e. those with sizes comfortably larger than indicated by
542: the curves in Figure~4, e.g. in the range $> 10-100$~m, may experience a slow enough radial
543: decay to allow the assembly of satellites in the disk.
544: However, addressing whether or not satellites can really form will require to include grain growth and the Stokes
545: drag regime, as well as self-consistent initial sizes and velocities for boulders entering the circumplanetary disk.
546:
547: \section{Conclusions}
548:
549: We have shown the disk thermodynamics has an impact on all the most important aspects of disk-planet interaction for
550: Jupiter-mass planets. The way migration and accretion are affected is qualitatively consistent with the findings
551: by \citet{Paardekooper06} for low-mass planets, although in that work the effect was quantitatively much stronger.
552:
553: Migration is slowed down relative to isothermal runs because heating modifies the relative strength of the inner and outer spiral arm,
554: or stifles the spiral perturbations completely as in the case of adiabatic runs. This effect partially counterbalances the faster
555: migration found in our isothermal disks relative to standard estimates that neglect viscosity, self-gravity and the motion of
556: the planet. If the slower migration seen in radiative transfer runs persists over $> 100$ orbits one would expect the mean
557: migration rate to differ from the estimates of standard type II migration by no more than a factor of 2, confirming the earlier
558: conclusions reached by d'Angelo et al. (2003) using 2D simulations.
559: The mass transport towards the planet is hampered by the increasing pressure gradient in non-isothermal runs. This will have
560: to be investigated further by including a proper accretion model in the simulations \citep{d'Angelo03,KK06}.
561:
562: %\citet{KK06} did not
563: %find clear differences when comparing migration in isothermal and radiative runs
564: %for Jupiter mass planets. Unfortunately, a detailed comparison with their work %\citet{KK06}
565: %is difficult. Nearly all their isothermal simulations start with a gap while all our runs, both isothermal
566: %and non-isothermal, have no initial gap. Gaps in our runs are shallower than theirs even after many orbits because of
567: %viscosity. In brief, our planets are always more coupled with the disk, which affects both the gas inflow towards the
568: %planet and the torques.
569:
570: The different structure of the circumplanetary envelope depending
571: on disk thermodynamics that we find is qualitatively
572: in agreement with the results of \citet{KK06}, although our disks with flux-limited diffusion are less sub-keplerian than theirs.
573: Satellite formation is not easy in the thick circumplanetary disk, but it may still be possible provided that the solid component
574: is dominated by large boulders.
575:
576: %, but we also find that this might happen if most of the solid material is in sizes ....
577: %Overall our results confirm that realistic thermodynamics, that were neglected completely until recently, are crucial to
578: %understand disk-planet interaction and predict correctly migration rates, planetary growth rates and the nature of the gas
579: %flow around the planet in the core-accretion model. In addition, this has to be coupled with a sufficiently high resolution
580: %if the objective is to determine the structure of the circumplanetary envelope.
581:
582: \section{Acknowledgments}
583:
584: The authors wish to thank H. Klahr, S.-J. Paardekooper and G. Mellema for fruitful discussions.
585: %L. F. thanks R. Feldman and S. Callegari for providing some post-processing programs.
586: The simulations were run on the GONZALES cluster at ETH Zurich and on the ZBOX2 supercomputer
587: at the University of Zurich.
588:
589: \begin{thebibliography}{}
590: \bibitem[Alexander et al.(2006)]{Alexander06} Alexander, R. D., Clarke, C. J., Pringle, J. E., 2006, \mnras, 369, 216
591: \bibitem[Alibert et al.(2005)]{Alibert05} Alibert, Y., Mordasini, C., Benz, W., Winisdoerffer, C., \aap, 434, 343
592: \bibitem[Balsara(1995)]{Balsara95} Balsara, D. S.,1995, Journal of Computational Physics, 121, 357
593: \bibitem[Baruteau \& Masset(2008)]{Baruteau08} Baruteau, C., Masset, F., 2008, \apj, 678, 483
594: \bibitem[Bate et al.(2003)]{Bate03} Bate, M. R., Lubow, S. H., Ogilvie, G. I., Miller, K. A., 2003, \mnras, 341, 213
595: \bibitem[Crida et al.(2007)]{Crida07} Crida, A., Morbidelli, A., Masset, F., 2007, \aap, 461, 1173
596: \bibitem[d'Alessio et al.(1997)]{d'Alessio97} d'Alessio, Paola; Calvet, Nuria; Hartmann, Lee, 1997, \apj, 474, 397
597: \bibitem[d'Angelo et al.(2003)]{d'Angelo03} d'Angelo, G., Henning, T., Kley, W., 2003, \apj, 599, 548
598: \bibitem[de Val-Borro et al.(2006)]{deValBorro06} de Val-Borro M. et al., 2006, \mnras, 370, 529
599: \bibitem[Edgar(2007)]{Edgar07}Edgar, R. G., 2007, \apj, 663, 1325
600: \bibitem[Ida \& Lin(2004)]{Ida04} Ida, S., Lin, D. N. C., \apj, 604, 388
601: %\bibitem[Klahr \& Kley(2006)]{KK06} Klahr, H., Kley, W, 2006, \aap, 445, 747
602: \bibitem[KK06()]{KK06} Klahr, H., Kley, W, 2006, \aap, 445, 747
603: \bibitem[Lin \& Papaloizou(1986)]{Lin86} Lin, D. N. C., Papaloizou, J. C. B., 1986, \apj, 309, 846
604: \bibitem[Lin \& Papaloizou(1993)]{Lin93} Lin, D. N. C., Papaloizou, J. C. B., 1993, prpl conf, 749
605: \bibitem[Lufkin et al.(2004)]{Lufkin04} Lufkin, G., Quinn, T., Wadsley, J., Stadel, J., Governato, F.,2004, \mnras, 347, 421
606: \bibitem[Masset \& Papaloizou(2003)]{Masset03} Masset, F. S., Papaloizou, J. C. B., \apj, 588, 494
607: \bibitem[Mayer et al.(2007)]{Mayer07} Mayer, L., Lufkin, G., Quinn, T., Wadsley, J., 2007, \apj, 661, 77
608: \bibitem[Nelson \& Benz(2003a)]{Nelson03a} Nelson, A. F., Benz, W., 2003a, \apj, 589, 556
609: \bibitem[Nelson \& Benz(2003b)]{Nelson03b} Nelson, A. F., Benz, W., 2003b, \apj, 589, 578
610: %\bibitem[Nelson(2005)]{Nelson05} Nelson, R. P., 2005, \aap, 443, 1067
611: %\bibitem[Paardekooper \& Mellema(2006)]{Paardekooper06} Paardekooper, S.-J., Mellema, G.,2006, \aap, 459, 17
612: %\bibitem[Paardekooper \& Mellema(2008)]{Paardekooper08} Paardekooper, S.-J., Mellema, G.,2008, \aap, 478, 245
613: \bibitem[PM06()]{Paardekooper06} Paardekooper, S.-J., Mellema, G.,2006, \aap, 459, 17
614: \bibitem[PM08()]{Paardekooper08} Paardekooper, S.-J., Mellema, G.,2008, \aap, 478, 245
615: \bibitem[Papaloizou \& Nelson(2003)]{Papaloizou03} Papaloizou, J. C. B., Nelson, R. P., 2003, MNRAS 339, 983
616: \bibitem[Papaloizou et al.(2007)]{PPV} Papaloizou, J. C. B., Nelson, R. P., Kley, W., Masset, F. S., Artymowicz, P., 2007, prpl conf, 655
617: \bibitem[Pollack et al.(1996)]{Pollack96} Pollack, J. B., Hubickyj, O., Bodenheimer, P., Lissauer, J. J., Podolak, M., Greenzweig, Y., Icarus, 124, 62
618: %\bibitem[Shapiro \& Teukolsky(1985)]{Shapiro85} Shapiro, S. L., Teukolsky, S. A., 1985, \apj, 298, 58
619: \bibitem[Wadsley et al.(2004)]{Wadsley04} Wadsley, J. W., Stadel, J.,
620: Quinn, T, 2004, New Astronomy, 9, 137
621: \bibitem[Ward(1997)]{Ward97} Ward, W. R., 1997, \apj, 482, 211
622: \bibitem[Weidenschilling(1977)]{Weidenschilling77} Weidenschilling, S. J., 1977, MNRAS, 180, 57
623: \end{thebibliography}
624: \end{document}