1: 

2: %

3: %\documentstyle[12pt,aaspp4]{article}

4: %\documentstyle[12pt,aasms4]{article}

5: %\documentstyle[emulateapj]{article}

6: 

7: %\documentclass[preprint]{aastex}

8: \documentclass[manuscript]{aastex}

9: %\usepackage{aastexug}

10: 

11: \def\gta{\gtrsim}

12: \def\lta{\lesssim}

13: \def\kms{{ {\rm \,km\,s^{-1}} }}

14: \def\ms{{M_*}}

15: \def\be { \begin{equation} }

16: \def\ee { \end{equation} }

17: \def\L {{\cal L}}

18: \def\LLp {{ {m_1\over\mp}\left({j_1\over\jp}\right)^{1/3}

19: \sqrt{1-e_1^2\over 1-\ep^2} }}

20: \def\jp {{j_p}}

21: \def\np {{n_p}}

22: \def\mp {{m_p}}

23: \def\ep {{e_p}}

24: \def\ap {{a_p}}

25: \def\Ep {{E_p}}

26: \def\Lp {{L_p}}

27: \def\Yp {{Y_p}}

28: \def\dEdEp {{1+{m_1\over \mp}\left({\jp\over j_1}\right)^{2/3} }}

29: 

30: \begin{document}

31: 

32: %\shortauthors{Murray et al.}

33: 

34: 

35: \title{\bf Eccentricity Evolution of Resonant Migrating Planets} 

36: \author{N.  Murray\altaffilmark{1},

37: M. Paskowitz, and M. Holman\altaffilmark{2}}

38: 

39: \altaffiltext{1}{Canadian Institute for Theoretical Astrophysics,

40:  60 St. George st., University of Toronto, Toronto, ONT M5S 3H8,

41: Canada; murray, paskowitz@cita.utoronto.ca}

42: 

43: 

44: \altaffiltext{2}{Harvard-Smithsonian Center for Astrophysics, 60

45: Garden Street, Cambridge, MA, 02138, USA; mholman@cfa.harvard.edu}

46: 

47: 

48: \begin{abstract}

49: We examine the eccentricity evolution of a system of two planets

50: locked in a mean motion resonance, in which the outer planet loses

51: energy and angular momentum. The sink of energy and angular momentum

52: could be either a gas or planetesimal disk. We show that the

53: eccentricity of both planetary bodies can grow to large values,

54: particularly if the inner body does not directly exchange energy or

55: angular momentum with the disk. We analytically calculate the

56: eccentricity damping rate in the case of a single planet migrating

57: through a planetesimal disk. We present the results of numerical

58: integrations of two resonant planets showing rapid growth of

59: eccentricity. We also present integrations in which a Jupiter-mass

60: planet is forced to migrate inward through a system of $5-10$ roughly

61: Earth mass planets. The migrating planet can eject or accrete the

62: smaller bodies; roughly $5\%$ of the mass (averaged over all the

63: integrations) accretes onto the central star. The results are

64: discussed in the context of the currently known extrasolar planetary

65: systems.

66: \end{abstract}

67: 

68: \keywords{planetary systems---stars: }

69: 

70: \section{INTRODUCTION}

71: The sixty or so extrasolar planetary systems known to date have

72: revealed three striking features (for an up to date list of systems

73: and their properties see http://www/exoplanets.org/ or

74: http://www.obspm.fr/encycl/encycl.html). First, the distribution of

75: orbital semimajor axes of the planets range from $\sim3$ AU down to an

76: almost incredible $0.038$ AU. Second, most of the objects have high

77: eccentricity by solar system standards, with a typical value being

78: around $e=0.4$, but ranging up to $0.927$. Third, the parent stars are

79: highly metal rich, and appear to have accreted iron rich

80: material after having reached the main sequence

81: \cite{gonzalez,santos,laughlin,murray}.

82: 

83: The simplest interpretation of the small orbits is that Jupiter-mass

84: planets experience large scale migrations in some cases, but not in

85: others; Jupiter falls into the latter class. There are currently two

86: viable explanations for the migration, tidal interactions between the

87: planet and the gas disk out of which it formed \cite{GT,lbr}, and gravitational

88: interactions between the planet and a massive (1-5 Jupiter mass)

89: planetesimal disk (Murray et al 1998).

90: 

91: The most straightforward interpretation of the high eccentricities,

92: that they result from collisions or near collisions of two or more

93: Jupiter-mass planets, is appealing, but require that most systems are

94: dynamically unstable, in addition to undergoing

95: migration. Furthermore, a recent exhaustive study of the problem

96: indicates that the number of systems with low eccentricities is

97: smaller than would be produced by collisions and scattering \cite{ford}.

98: 

99: In this paper we investigate another possible mechanism for producing

100: large eccentricities; resonant migration. We suppose that a

101: Jupiter-mass planet is forced to migrate inward, either by tidal

102: torques or by ejection of planetesimals, and that a second (possibly

103: much less massive object) is in a mean motion resonance with the

104: first. We further assume that the migration process does not

105: significantly damp the eccentricity of the inner body. This could

106: occur in migration in a gas disk if the inner disk manages to drain

107: onto the central star while leaving behind the planets and a

108: substantial outer gas disk. It would almost inevitably occur in

109: migration through a massive planetesimal disk, since the planetesimals

110: are likely to accrete into terrestrial mass or larger bodies; we show

111: below by direct numerical integrations that these $1-50M_\oplus$

112: bodies will be trapped into mean motion resonances.

113: 

114: We show that the inward migration of two planets trapped in a mean

115: motion resonance can produce eccentricities as high as $0.7$. We also

116: show that in the case of migration by planetesimal ejection, that the

117: final state may or may not have two resonant planets. Whether the

118: distribution of eccentricity with planetary mass and semimajor axis

119: produced by such resonant migrations is consistent with the observed

120: distribution is a question left for later work.

121: 

122: As a byproduct of our numerical simulations, we find that the fraction

123: of planetesimal disk mass that accretes onto the star is likely to be much

124: smaller than found in the work of \cite{qh}; that work studied

125: the accretion of massless test particles subject to gravitational

126: perturbations from a migrating Jovian-mass planet. The authors found

127: that of order half the mass in the disk would accrete. Using our more

128: realistic, but less extensive integrations of massive planetesimals,

129: we find a much smaller fraction ($\sim5\% $) of the disk mass

130: accreting onto the star in the early stages of the migration. (Another

131: $\sim5-10\%$ of the disk mass will fall on the star if the planet

132: approaches within $\sim0.1$ AU)\cite{hansen}.

133: 

134: The remainder of the paper is organized as follows. In contrast to

135: tidal torque migration the eccentricity evolution of a Jupiter-mass

136: object migrating through a massive planetesimal disk has not been

137: extensively studied. Section 2 gives a short derivation in the case

138: that the migrating planet does not accrete a substantial fraction of the

139: planetesimals. This is appropriate when the escape velocity from the

140: surface of the planet is larger than the escape velocity at the

141: orbital distance of the planet from the star. Section 2 describes the

142: process of capture into resonance, and the evolution of the

143: eccentricities of both resonant bodies as the migration

144: proceeds. Section 4 presents the results of numerical integrations of

145: two resonant bodies, with parameters appropriate for planetesimal

146: migrations, as well as integrations involving up to $11$

147: planets. Section 5 gives a discussion of our results, and contrasts

148: the two types of migration, while section 6 presents our conclusion.

149: 

150: \section{MIGRATION AND ECCENTRICITY EVOLUTION}

151: We examine the eccentricity evolution of a

152: Jupiter-mass body migrating inwards due to the extraction of energy

153: and angular momentum.  The energy $\Ep$ and angular

154: momentum $\Lp$ of the planet are given by

155: %

156: \begin{eqnarray} %$

157: &\Lp=m_p\sqrt{GM_*a_p(1-\ep^2)}\label{ELL}\label{energy}\\

158: & \Ep=-{GM_*m_p\over 2a_p}\label{ELE} 

159: \end{eqnarray} %$

160: %

161: Taking the time derivative of equation (\ref{energy}), we find the

162: time variation of $\ep$ in terms of the time variation of the

163: planetary energy, assuming the planetary mass is

164: fixed:

165: %

166: \be \label{ecc_evol}%$

167: {\ep\over 1-\ep^2}{d\ep\over dt}

168: =-{1\over2}\left({1\over\Ep}{d\Ep\over dt}\right)

169: \left[1+2\left({\Ep\over \Lp}{d\Lp\over d\Ep}\right)\right].

170: \ee %$

171: %

172: The quantity 

173: %

174: \be \label{beta}%$

175: \beta\equiv\left[1+2\left({\Ep\over \Lp}{d\Lp\over d\Ep}\right)\right]

176: \ee %$

177: %

178: is a convenient measure of the rate at which the planetary

179: eccentricity changes.  Both $E_p$ and $dE_p/dt$ are negative for an

180: inward migration, so $e_p$ decreases if $\beta>0$. Conservation of

181: energy and angular momentum implies that $d\Lp/d\Ep=(dL/dE)_T$, where

182: the latter quantity is the ratio of the rates at which angular

183: momentum and energy are removed from the system by whatever process is

184: driving the migration.  The planetary eccentricity decreases as long

185: as

186: %

187: \be \label{damping}%$

188: \left({dL\over dE}\right)_T<-{\Lp\over2\Ep}=\sqrt{1-\ep^2}\Big/n_p.

189: \ee %$

190: %

191: 

192: We now specialize to the case of planetesimal migration.  To find

193: $dL_p/dE_p$, we calculate the total change in $E$ and $L$ for a

194: planetesimal of mass $m$ (where $m<<m_p$) from its initial orbit, with

195: semimajor axis $a$ and eccentricity $e$, to the point at which it is

196: ejected, then use conservation of energy and angular momentum. We note

197: that this is not adequate for cases where the planet eats the

198: planetesimal, since some orbital energy will be lost in the form of

199: radiation in that case.

200: 

201: Figure (\ref{Fig_EL}) illustrates the constraints on the evolution of

202: a planetesimal in the $E-L$ plane, in the case $e_p=0.1$. The solid

203: curved line corresponds to $e=0$; planetesimal orbits must lie to the

204: left of this line. In analyzing motion involving close encounters

205: with a planet, it is useful to introduce the Jacoby parameter,

206: $J=E-n_pL$, where $n_p$ is the mean motion of the planet. The diagonal

207: solid line corresponds to $J/m=-3/2$ (in units where $GM_*=\ap=1$). We

208: note that in order to be ejected from $a<\ap$ the asteroid must reach

209: $J/m\ge -3/2$, since it must pass through $a=a_p$ with $e\ge0$ to be

210: ejected.

211: 

212: We will assume that the planetesimal is ejected with zero total

213: energy; if it is ejected with a larger energy, the damping rate will

214: be smaller than the estimate we obtain below. The initial energy and

215: momentum of the planetesimal are given by expressions analogous to

216: equations (\ref{ELL}) and (\ref{ELE}).  Rather than calculating the

217: change in $L$ directly, we calculate the change in the Jacoby

218: parameter; we do so because $J$ is constant (in a

219: statistical sense) during the planet crossing phase of the asteroid's

220: evolution \cite{opik}. To lowest order in $m/m_p$ we have

221: %

222: \be \label{dlde}%$

223: \left({dL\over dE}\right)_T=\left[1-{dJ\over dE}\right]\Bigg/n_p.

224: \ee %$

225: %

226: If the planetesimal disk is originally cold ($e<<1$) and $e_p<<1$, few

227: planetesimals will cross the orbit of the planet. However,

228: planetesimals trapped in resonance with the planet will suffer chaotic

229: perturbations which on average transfer angular momentum, but not

230: energy, from the asteroid to the planet. This causes $J$ to increase

231: while leaving $E$ fixed. Once enough angular momentum has been removed

232: from the asteroid's orbit, the asteroid can suffer close encounters

233: with the planet. We assume that the first close encounter removes the

234: asteroid from the resonance, while leaving $J$ fixed. Subsequent

235: encounters extract or supply energy and angular momentum to the

236: asteroid in such a way as to leave $J$ constant on average, as noted

237: above. Eventually the planetesimal is ejected with $E\ge0$; taking

238: $E=0$ we find

239: %

240: \be \label{djde1}%$

241: {dJ\over dE}=2\left({a\over \ap}\right)^{3/2}

242: \left(\sqrt{1-e^2}-\sqrt{1-e_c^2}\right),

243: \ee %$

244: %

245: where $e_c\equiv \ap(1-\ep)/a-1$ is the eccentricity at which the

246: planetesimal just crosses the orbit of the planet. Note that $dJ/dE\ge0$.

247: 

248: In arriving at equation (\ref{djde1}) we have assumed that the final

249: Jacoby parameter 

250: %

251: \be %$

252: J_f/m=J_c/m\equiv-{1\over 2a}-\sqrt{a(1-e_c^2)}>-3/2

253: \ee %$

254: %

255: (we again use $GM_*=\ap=1$). If the Jacoby parameter at planet

256: crossing ($J_c$) is not larger than $-3m/2$, the planetesimal must

257: diffuse to higher $J$ in order to be ejected, since it has to get past the

258: solid curve in Figure (\ref{Fig_EL}). Setting $J_f/m=-1.5$ we find

259: %

260: \be \label{djde2}%$

261: {dJ\over dE}=2\left({a\over \ap}\right)^{3/2}

262: \left[\sqrt{1-e^2}+{1\over 2}\left({\ap\over a}\right)^{3/2}-{3\over2}

263: \left({\ap\over a}\right)^{1/2}\right].

264: \ee %$

265: %

266: When $J_c/m>-3/2$ equation (\ref{djde1}) should be used, while equation

267: (\ref{djde2}) is appropriate if $J_c/m<-3/2$.

268: 

269: Combining equations (\ref{ecc_evol}) and (\ref{dlde}), the expression

270: for the rate of change of the planet's eccentricity is

271: %

272: \be \label{planetesimal_damping}%$

273: {\ep\over1-\ep^2}{d\ep\over dt}=-{1\over2}

274: \left({1\over\Ep}{d\Ep\over dt}\right)

275: \left[1-{\left(1-dJ/dE\right)\over \sqrt{1-\ep^2}}\right]

276: \ee %$

277: %

278: 

279: It can be shown, using equations (\ref{djde1}) and (\ref{djde2}), that

280: %

281: \be %$

282: {dJ\over dE}\ge1-\sqrt{1-\ep^2}.

283: \ee %$

284: %

285: In other words, planetesimal migration, as described here, always

286: damps the eccentricity of a single planet. We can estimate the value

287: of $dJ/dE$ when $e$ and $\ep$ are small; for example for $J_c<-3/2$

288: (which requires $e\lta (3+2\sqrt{3})\ep$) equation

289: (\ref{djde2}) evaluated at the maximum $a/\ap=(1-\ep)/(1+e)$ gives

290: %

291: \be %$

292: {dJ\over dE}\approx {3\over4}\ep^2+{3\over 2}\ep e-{1\over 4}e^2.

293: \ee %$

294: %

295: For $e\approx \ep=0.05$, $dJ/dE\approx 2\ep^2\approx0.005$. For

296: smaller values of $a$ this increases, as shown in Figure

297: \ref{Fig_dJdE}. Panel (a) in the Figure shows $\beta$ and $dJ/dE$ for

298: a planet with $\ep=0.05$ ejecting a planetesimal with initial

299: $e=0.05$, as a function of the initial $a$ of the planetesimal. The

300: relevant value of $dJ/dE$ depends on the average initial $a$ of the

301: planetesimals that are ejected, i.e., it depends on which resonance is

302: most actively ejecting objects. Early on in the evolution of the

303: system we expect the planetesimal disk to be truncated inside the

304: chaotic zone produced by the overlap of first order mean motion

305: resonances (the ``$\mu^{2/7}$ chaotic region''\cite{wisdom}). Under

306: those circumstances the relevant resonance is the $5/3$, at

307: $a/\ap\approx 0.71$; $dJ/dE\approx 0.066$ for that case. However, as

308: the migration proceeds, material is supplied to the $\mu^{2/7}$ zone,

309: and $dJ/dE$ will be on average smaller, at least while $e_p<<1$. As

310: $e_p$ increases, equations (\ref{djde1}) and (\ref{djde2}) show that

311: $dJ/dE$ no longer increases as rapidly as $e_p^2$; it effectively

312: saturates near $dJ/dE\approx0.3-0.4$. Panel (b) shows $\beta$ and

313: $dJ/dE$ in the case $\ep=e=0.5$. Planetesimals never get the chance to reach

314: the $5/3$ resonance, since they become planet crossing at much smaller

315: semimajor axis. The damping rate is of order $0.35$. 

316: 

317: We can calculate the circularization time for planetesimal migration,

318: in terms of the migration time. Following the notation employed in

319: satellite studies \cite{lpc},

320: %

321: \begin{eqnarray} %$

322: {d\Ep\over dt}&=&-(\np T+H)\label{torque}\\

323: {d\Lp\over dt}&=&-T,

324: \end{eqnarray} %$

325: %

326: where $T$ is the (average) torque exerted on the planet by the

327: ejection of planetesimals, and $H$ is responsible for removing energy

328: from the radial motion of the planet, i.e., it damps the eccentricity.

329: (Actually it would be better to use

330: $d\Ep/dt=-(\np T/\sqrt{1-\ep^2}+H)$, as will become apparent). It

331: follows from these equations that

332: %

333: \be %$

334: H=-\beta{d\Ep\over dt},

335: \ee %$

336: %

337: where we have replaced the missing factor of

338: $\sqrt{1-\ep^2}$. Using this in equation (\ref{torque})

339: we define the migration time 

340: %

341: \be %$

342: \tau_m\equiv {GM_*\mp\over 2\ap\np T}(1-\beta)

343: \ee %$

344: %

345: 

346: The circularization time $\tau_c$ is given by equation (\ref{ecc_evol});

347: %

348: \be %$

349: \tau_c\equiv \tau_m{\ep^2\over \beta(1-\ep^2)}.

350: \ee %$

351: %

352: For small $\ep$ this reduces to $\tau_c\approx(2/3)\tau_m$; the

353: circularization time is comparable to the migration time.

354: 

355: Note that in deriving equations (\ref{djde1}) and (\ref{djde2}) we

356: have ignored the finite extent $D\equiv(m_p/3M_*)^{1/3}a_p$ of the

357: planet's Hill sphere, the region over which the planet's gravity

358: exceeds the tidal acceleration from the central star. For a Jupiter

359: mass planet $D\approx 0.07a_p$. Including the effect of the Hill

360: sphere in our analysis effectively increase $\ep$ to

361: $\ep'=\ep+(\mp/3M_*)^{1/3}$; this is why we use $\ep=0.1$ in Figure

362: \ref{Fig_EL}.

363: 

364: 

365: \section{RESONANCE CAPTURE AND ECCENTRICITY EVOLUTION}

366: We have seen that a single planet embedded in a planetesimal disk

367: suffers eccentricity damping (it appears that a similar statement

368: applies to a single planet in a gas disk \cite{PNM}). However, a

369: Jupiter mass planet migrating through a disk of planetesimals will

370: capture bodies into resonance; in the early stages this is how the

371: migration proceeds. We show in this section that these resonant bodies

372: tend to increase the eccentricity of the Jupiter mass planet; if

373: $10-20$ Earth masses (denoted $M_\oplus$) are trapped into a

374: resonance, then this resonant eccentricity driving exceeds the

375: eccentricity damping described in the previous section, and the

376: eccentricity of the planet will increase as it migrates inward.

377: 

378: We begin by describing capture into resonance. We consider the

379: gravitational interaction of two planets in orbit around a much more

380: massive central body. For simplicity we consider only the planar

381: problem.  In the absence of dissipative effects the Hamiltonian

382: describing the motion is

383: %

384: \be  \label{Hamiltonian}%$

385: H=-{\mu_1^2m_1\over 2L_1^2}-{\mu_p^2\mp\over 2\Lp^2}-

386: {Gm_1\mp\over a_1}\sum_{\bf j}\Phi_{\bf j}(a_1,\ap)e_1^{|j_3|}\ep^{|j_4|}

387: \cos\left[j_1\lambda_1-\jp\lambda_p+j_3\varpi_1+j_4\varpi_p

388: \right].

389: \ee  %$

390: %

391: Here $\mu_1\equiv{\cal G}(\ms+m_1)$, where ${\cal G}$ is Newton's

392: gravitational constant, and $L_1=\sqrt{\mu_1 a_1}$, with similar

393: definitions for the outer planet (labeled with a subscript $p$).  The

394: third term in equation (\ref{Hamiltonian}) represents the mutual

395: perturbations of the two planets. It produces variations in the

396: orbital elements ($a$, $e$, and so forth) of order the planetary mass

397: $m_i$.

398: 

399: The coefficient $\Phi\sim \left[a_1/(\ap-a_1)\right]^{|j_1-\jp|}$

400: \cite{hm}. The integers $j_i$ satisfy the relation

401: $j_1-\jp+j_3+j_4=0$. Each cosine term in the sum is referred to as a

402: resonant term or simply as a resonance. Resonances with $|j_1-\jp|=q$

403: are proportional to $q$ powers of eccentricity, and are said to be

404: $q$th order mean motion resonances.  The planets are said to be in

405: resonance if one or more of the arguments of the cosines are

406: bounded. Since $n_1\equiv\left<\dot\lambda_1\right>$ (where the angle

407: brackets refer to an average over a single orbit) and $\np$ are much

408: larger than $\dot\varpi_1$ and $\dot\varpi_p$, the condition for resonance

409: is roughly equivalent to

410: %

411: \be \label{lock}%$

412: j_1n_1-\jp\np=0, 

413: \ee %$

414: %

415: or $\ap/a_1=(\jp/j_1)^{3/2}$. Throughout this section we ignore

416: non-resonant terms of second order in the planetary masses.

417: 

418: Suppose that $\ap/a_1$ is initially larger than this resonant value,

419: but that some dissipative process acts to reduce $\ap$ while leaving

420: $a_1$ unchanged. Then the torques represented by the resonant cosine

421: term will increase, since $\Delta a\equiv \ap-a_1$ is decreasing and

422: the torques are proportional to $(a_1/\Delta a)^2$. Another way to say

423: this is that the depth of the potential well represented by the

424: resonant term is increasing, as is the width of the resonance. As

425: $\ap/a_1$ passes through the resonant value $(j_1/\jp)^{2/3}$, the

426: planets may be trapped into resonance. The torques represented by the

427: resonant term in eqn. (\ref{Hamiltonian}) will then transfer energy

428: and angular momentum between the two planets in just such a way as to

429: maintain the resonance while both bodies move toward the star

430: \cite{goldreich}.

431: 

432: Capture is much less likely if the planets are moving away from each

433: other, for example if $a_1$ is decreasing, or if both semimajor axes

434: are decreasing but that of the inner planet decreases more rapidly; in

435: that case the size of the resonance is decreasing, and the planets

436: will usually pass through the resonance without being trapped.

437: 

438: Henceforth we will assume that the outer planet is moving toward the

439: inner planet, resulting in capture into resonance. 

440: 

441: Suppose for the moment that the inner planet has a sufficiently small

442: mass that it cannot effectively scatter planetesimals, so that it does

443: not lose energy or angular momentum directly to the planetesimal

444: bath. (In the case of gas migration, we assume that the inner disk is

445: non-existent). By virtue of its resonance interaction with the outer planet, it

446: nevertheless does supply energy and angular momentum indirectly to

447: material driving the migration. Another way to

448: say this is that $dE_p/dL_p$ no longer equals $(dE/dL)_T$, the quantity

449: calculated for the case of planetesimal migration in the previous section.

450: Here we calculate the relation between these two quantities that

451: obtains when a second object of mass $m_1$, is in

452: resonance with the Jupiter-mass object. 

453: 

454: We relate $dE_p/dL_p$ to $(dE/dL)_T$ using the conservation of energy

455: and angular momentum. Conservation of energy gives

456: %

457: \be %$

458: {dE\over d\Ep}=1+{m_1\over \mp}\left({\jp\over j_1}\right)^{2/3},

459: \ee %$

460: %

461: while conservation of angular momentum implies

462: %

463: \be %$

464: {d\Lp\over d\Ep}=\left({dL\over

465: dE}\right)_T\left[\dEdEp\right]-{dL_1\over dE_p}

466: \ee %$

467: %

468: We also need

469: %

470: \be %$

471: {L_1\over\Lp}=\LLp.

472: \ee %$

473: %

474: Using all these in (\ref{ecc_evol}) we find

475: %

476: \be %$

477: {\ep\over1-\ep^2}{d\ep\over dt}=-{1\over2}

478: \left({1\over\Ep}{d\Ep\over dt}\right)

479: \left[

480: 1+2{\Ep\over\Lp}\left( {dL\over dE} \right)_T \left( {dE\over d\Ep}\right)

481: \right] 

482: +{L_1\over\Lp}\left({1\over L_1}{dL_1\over dt}\right).

483: \ee %$

484: %

485: Note that the $dE/d\Ep$ factor multiplying $(dL/dE)_T$ is larger than

486: one; it effectively increases $(dL/dE)_T$. From equation

487: (\ref{damping}) we see that this will tend to increase $\ep$. The term

488: proportional to $dL_1/dt$ will tend to decrease $\ep$, but we shall

489: see that its effect is smaller than that of the term involving

490: $dE/d\Ep$; this is the origin of the increase in eccentricity of two

491: resonant bodies undergoing migration.  From the expression for $L_1$,

492: and using the resonance condition, we have

493: %

494: \be %$

495: \left({1\over L_1}{dL_1\over dt}\right)=

496: -{1\over 2}\left({1\over\Ep}{d\Ep\over dt}\right)

497: -{e_1\over 1-e_1^2}{de_1\over dt}.

498: \ee %$

499: %

500: As just noted the first term on the right will tend to damp the

501: eccentricity of the outer planet; the second term on the right will

502: also damp $\ep$ as long as $de_1/dt>0$. Combining the last two

503: equations we find

504: %

505: %%\begin{eqnarray} %$

506: %%{\ep\over1-\ep^2}{d\ep\over dt}&=&-{1\over2}

507: %%\left({1\over\Ep}{d\Ep\over dt}\right)

508: %%\left[

509: %%1+2{\Ep\over\Lp}\left( {dL\over dE} \right)_T\left( {dE\over d\Ep}\right)

510: %%+{m_1\over\mp}\left({j_1\over \jp}\right)^{1/3}\sqrt{1-e_1^2\over 1-\ep^2}

511: %%\right] \nonumber \\

512: %%&&-{m_1\over\mp}\left({j_1\over \jp}\right)^{1/3}\sqrt{1-e_1^2\over

513: %%1-\ep^2}

514: %%{e_1\over 1-e_1^2}{de_1\over dt}.

515: %%\end{eqnarray} %$

516: %

517: %%Now add and subtract $(m_1/\mp)(\jp/j_1)^{2/3}$ inside the square

518: %%brackets to get

519: %

520: \begin{eqnarray} \label{almost_there}%$

521: {\ep\over1-\ep^2}{d\ep\over dt}&=&-{1\over2}

522: \left({1\over\Ep}{d\Ep\over dt}\right)

523: \Bigg\{

524: \left[1+2{\Ep\over\Lp}\left( {dL\over dE} \right)_T\right]

525: \left( {dE\over d\Ep}\right) \nonumber\\

526: &&-{m_1\over\mp}\left({j_1\over \jp}\right)^{1/3}{1\over\sqrt{1-\ep^2}}

527: \left[

528: \left({\jp\over j_1}\right)\sqrt{1-\ep^2}-\sqrt{1-e_1^2}

529: \right]

530: \Bigg\} \nonumber\\

531: &&-{m_1\over\mp}\left({j_1\over \jp}\right)^{1/3}\sqrt{1-e_1^2\over

532: 1-\ep^2}

533: {e_1\over 1-e_1^2}{de_1\over dt},

534: \end{eqnarray} %$

535: %

536: where we have added and subtracted $(m_1/\mp)(\jp/j_1)^{2/3}$ inside

537: the curly brackets.

538: 

539: We identify 

540: %

541: \be %$

542: \left({d\Ep\over dt}\right)_T\equiv\left({d\Ep\over dt}\right)\left( {dE\over d\Ep}\right)

543: \ee %$

544: %

545: as the rate at which the expulsion of planetesimals removes energy

546: from the outer planet. With this identification, and recalling

547: equation (\ref{ecc_evol}), it becomes clear that

548: the first term in curly brackets in equation (\ref{almost_there}) is

549: the time rate of change of $\ep$ due to the expulsion of

550: planetesimals. The final result is

551: %

552: \begin{eqnarray} %$

553: {\ep\over1-\ep^2}{d\ep\over dt}&=& 

554: -{1\over2}\left({1\over \Ep}{d\Ep\over dt}\right)_T

555: \Bigg\{

556: 1-{(1-dJ/dE)\over \sqrt{1-\ep^2}}\nonumber\\

557: &&-{m_1\over\mp}\left({j_1\over \jp}\right)^{1/3}{1\over\sqrt{1-\ep^2}}

558: \left[

559: \left({\jp\over j_1}\right)\sqrt{1-\ep^2}-\sqrt{1-e_1^2}

560: \right]\Bigg/\left[\dEdEp\right]

561: \Bigg\} \nonumber\\

562: &&-\LLp{e_1\over1-e_1^2}{de_1\over dt}\label{ep}

563: \end{eqnarray} %$

564: %

565: Note that this expression reduces to equation

566: (\ref{planetesimal_damping}) when $m_1\to0$.

567: 

568: Equation (\ref{ep}) has two undetermined quantities ($d\ep/dt$ and

569: $de_1/dt$). Bodies that are trapped in a mean motion resonance typically have

570: their apsidal lines locked as well, so that

571: $\dot\varpi_1=\dot\varpi_2$. Using the equations of motion for

572: $\varpi_1$ and $\varpi_2$, we can find a relation between $e_1$ and

573: $\ep$ that depends on the precession rates of the apsidal lines. The

574: latter are determined both by the mutual perturbations of the two

575: planets, and by the distribution of mass in the planetesimal (or gas)

576: disk. Given the current state of both observations and theory, we feel

577: that a detailed calculation is not justified.

578: 

579: In the appendix we present another derivation in which we allow for

580: the possibility that the inner, less massive planet also loses energy

581: and angular momentum to the sink of energy and angular momentum.

582: 

583: We proceed to examine some limiting cases. First, suppose that no

584: tides act on the inner planet, and that the tides acting on the

585: outer planet keep that planet's orbit circular. Then

586: %

587: \be %$ 

588: {e_1\over \sqrt{1-e_1^2}}{de_1\over dt}\approx

589: -\left({d\ln \ap\over dt}\right)_T 

590: \left[

591: {\jp\over j_1}-\sqrt{1-e_1^2}

592: \right]\Big/2\left(dE/d\Ep\right).

593: \ee %$

594: %

595: Since $\ap$ is decreasing, the right hand side is positive, and $e_1$

596: will grow. Assuming $\jp/j_1>>1$, we find

597: %

598: \be \label{e_1_of_a}%$

599: e_1\approx\sqrt{1-(1-\gamma_1\ln a_{1,i}/a_1)},

600: \ee %$

601: %

602: where $a_{1,i}$ is the semimajor axis of the inner body when it is

603: captured into resonance,  $\gamma_1=\jp/2j_1$, and we have neglected

604: the initial value of $e_1$.

605: 

606: Now suppose that no tides act on the inner body, and that the

607: migration has proceeded far enough that $e_1$ has grown to the point

608: that $de_1/dt$ is small. Then we can find the equilibrium value for

609: $\ep$ by setting the terms in the curly brackets equal, assuming that

610: $\beta=2\ep^2/3$;

611: %

612: \be \label{emax}%$

613: e_{p,max}\approx\sqrt{2m_1\ap\over 3(\mp a_1+m_1\ap)},

614: \ee %$

615: %

616: where we neglect $\sqrt{(1-e_1^2)/(1-\ep^2)}$ compared to

617: $\jp/j_1$. For gas-disk migration, the ratio of $\tau_c/\tau_m$ would

618: enter in the expression for $\beta$. For two equal mass bodies in a

619: $4/1$ resonance this is about $0.7$. Even for $m_1/\mp=0.1$ (a

620: $30M_\oplus$ inner planet) the equilibrium eccentricity is

621: $0.41$. However, note that this expression is actually an

622: underestimate in the case of planetesimal migration, since $\beta$,

623: which is a measure of the damping due to the migration process, does

624: not scale as $\ep^2$ for $\ep$ as large as $0.4$; $\beta$ actually

625: grows less rapidly, meaning that the damping is not as efficient as

626: equation (\ref{emax}) assumes.

627: 

628: We can relate the eccentricity to the distance migrated when

629: $\mp>>m_1$. Define

630: %

631: \be %$

632: \gamma\equiv{m_1\over \mp}\left({\jp\over

633: j_1}\right)^{2/3}.

634: \ee %$

635: %

636: 

637: The evolution of the eccentricity is then described by

638: %

639: \be \label{evolution}%$

640: {1\over 1-\ep^2}{d\ep^2\over dt}\approx 

641: -{1\over2}\left({1\over\Ep}{d\Ep\over dt}\right)_T

642: \left[\beta-\gamma\right]

643: \ee %$

644: % 

645: We have assumed that $\jp/j_1>>\sqrt{1-e_1^2}$, and neglected terms

646: second order in $m_1/\mp$. The eccentricity of the outer planet will

647: grow indefinitely providing $ \gamma>\beta$.

648: 

649: 

650: Now suppose that $\beta$ is independent of $\ep$; this is not true

651: when $\ep$ is small, but for small $\ep$ $\gamma$ can be much larger

652: than $\beta$; for $\ep\approx0.2$ or larger $\beta$ is

653: roughly constant. In that case we can integrate equation

654: (\ref{evolution}) to find the final eccentricity $e_{2f}$ of the outer

655: body,

656: %

657: \be \label{e_2_of_a}%$

658: e_{2f}\approx\sqrt{1-\left({a_{2f}\over a_{2i}}\right)^{\gamma-\beta}},

659: \ee %$

660: %

661: where $a_{2i}$ is the semimajor axis and eccentricity of

662: the outer planet when it enters the resonance, and $a_{2f}$ is the

663: semimajor axis when the planet either stops migrating or leaves the

664: resonance. We have assumed that the initial eccentricity $\ep<<1$; if it

665: is not, the final eccentricity will be larger.

666: 

667: The numerical work described below shows that a Jupiter-mass object

668: can capture smaller bodies into resonances ranging from the $2/1$ to

669: the $4/1$; we have even seen captures into the $11/2$ resonance. In a

670: scenario where the Jupiter-mass body migrates through a planetesimal

671: disk having a comparable mass, we expect resonance capture of

672: terrestrial bodies with masses ranging from 1 Earth mass ($M_\oplus$)

673: up to $30M_\oplus$ or more; an extreme upper limit might be of order

674: $50M_\oplus$, corresponding to about $10\%$ of the disk mass. The

675: plausible range for $\gamma$ is then $5\times10^{-3}-0.4$.  For

676: migration in a planetesimal disk we expect $\beta$ to be in the range

677: $0.01$ (for $\ep<<1$) to $0.3$ (for $\ep\gta0.5$).  Taking

678: $1$ AU as a representative value for $a_{2f}$ (although some

679: extrasolar planets are in much smaller orbits), with $a_{1f}\sim5-10$

680: AU, we find final eccentricities in the range $0.1-0.6$, with

681: $\ep\approx0.45$ being a typical value.

682: 

683: 

684: 

685: \section{Numerical Results}

686: We employ a Bulirsch-Stoer integrator with a variable step size. We

687: require that at each time step the relative accuracy of the

688: integration (as measured in phase space) be $10^{-12}$. Typical

689: orbital times are of order one to ten years, while the integrations

690: can extend up to $10^8$ yrs. In test runs where no energy is removed

691: from the planet, the largest variation in total energy is typically

692: less than a part in $10^9$. In most of the runs reported on here we

693: remove energy and angular momentum from the largest planet; the

694: variations in energy and angular momentum from the expected amounts

695: are similarly small.

696: 

697: In runs with multiple massive planets collisions often occur. We

698: assume that the smaller planets are rocky bodies with bulk densities

699: of $3\ {\rm g/cm^3}$, that collision occur when planets are within two

700: times the sum of the planetary radii (to allow for capture by tidal

701: disruptions) and that the captures are completely inelastic with no

702: loss of mass to small fragments. This is reasonable for collisions

703: involving the more massive objects.

704: 

705: We assume that the most massive body (``Jupiter'') migrates inward by

706: ejecting numerous planetesimals from the system \cite{mhht}. We

707: simulate this by extracting energy and angular momentum from the orbit

708: of Jupiter. We implement this numerically as follows.

709: 

710: The energy and angular momentum of the planet are given by

711: %

712: \be %$

713: E={1\over2}\mp (v_x^2+v_y^2)-{GM_*\mp\over r}

714: \ee %$

715: %

716: and 

717: %

718: \be %$

719: L=\mp(xv_y-yv_x)

720: \ee %$

721: %

722: where $v_x$ and $v_y$ are the velocity of the planet (we suppress the

723: subscript $p$ for ease of reading). Recall that we assume the planet

724: and planetesimal are coplanar. Taking time derivatives, we invert to find

725: %

726: \begin{eqnarray} \label{drag}%$

727: {dv_x\over dt}&=&\left({1\over \Ep}{d\Ep\over dt}\right)

728: {\Ep\over \mp {\bf r}\cdot{\bf v}}

729: \left[x+{v_y\over \np}\sqrt{1-\ep^2}(\beta-1)\right]\nonumber\\

730: {dv_y\over dt}&=&\left({1\over \Ep}{d\Ep\over dt}\right)

731: {\Ep\over \mp {\bf r}\cdot{\bf v}}

732: \left[y-{v_x\over\np}\sqrt{1-\ep^2}(\beta-1)\right],

733: \end{eqnarray} %$

734: %

735: where we have used equation (\ref{beta}).

736: We assume that the close

737: encounters that lead to changes in the planetesimal's orbital elements

738: occur on times much shorter than the orbital period, so that we can

739: take the position of the planet to be fixed.

740: 

741: Using these equations for numerical work is problematic, since the

742: vector dot product in the denominator vanishes at peri- and

743: apoapse. We regularize the equations by multiplying the right-hand

744: side by $2\sin^2f$, where $f$ is the true anomaly.

745: 

746: The terrestrial mass objects in our simulations do not have high enough

747: escape velocities to efficiently eject smaller bodies, so we do not

748: force them to migrate.

749: 

750: We have integrated the equations of motion for the two body problem

751: (the star and a massive planet) modified to account for the drag

752: imposed by the ejection of planetesimals, as given in equation

753: (\ref{drag}), regularized as noted above. The eccentricity $e_p$ and

754: semimajor axis $a_p$ decay as expected. We discuss the behavior of

755: systems with more than one massive planet in the following subsections.

756: 

757: \subsection{TWO MASSIVE BODIES \label{double}}

758: To test the basic idea that migration of two resonant bodies will

759: induce the growth of eccentricity, we have started two planets just

760: outside resonance, and applied the ``tides'' described in the previous

761: paragraphs. An example is shown in Figure (\ref{Fig_basic}). The inner

762: body has a mass of $20M_\oplus$ and the outer body has a mass equal to

763: that of Jupiter ($\approx 318M_\oplus$). The upper plot shows the

764: semimajor axes of both bodies as a function of time, while the lower

765: plot shows the eccentricities. Energy and angular momentum were

766: removed only from the outer body. The resonance interaction forces the

767: inner body to migrate inward as well, as can be seen in the

768: figure. One can also see from the figure that the eccentricity of both

769: bodies increased.

770: 

771: The prediction based on (\ref{e_2_of_a}) for the eccentricity of the

772: outer, more massive body is too large for $\ap$ near the initial

773: value, which is consistent with the fact that we neglected the rather

774: rapid variation of $e_1$; as the migration proceeds the prediction

775: becomes more accurate. By the end of the integration equation

776: (\ref{e_2_of_a}) is a good approximation.

777: 

778: 

779: \subsection{MULTIPLE MASSIVE BODIES\label{multiple}}

780: The planetesimal disk we postulate is very massive, $1-3$ Jupiter

781: masses. It is likely that multiply bodies with masses comparable to or

782: larger than that of the Earth are likely to form in such a massive

783: disk. As a first step toward a realistic simulation of a migration in

784: such a disk, we have run a number of cases involving five to ten

785: roughly Earth mass bodies placed on orbits with random semimajor axes

786: and small eccentricities inside the orbit of a Jupiter-mass planet. We

787: then force the Jupiter mass body to migrate inward toward the Earth

788: mass planets.

789: 

790: Figure (\ref{Fig_crash}) shows the result of one such integration. We

791: started five bodies with masses randomly distributed between $0.3$ and

792: $10M_\oplus$ with semimajor axes between $0.5$ and $4$ AU. As Jupiter

793: migrated inward, three of the small planets merged to form a

794: $6.3M_\oplus$ planet, one small planet crashed into Jupiter, and one

795: small planet was ejected. Both the latter two events illustrate the

796: migration mechanism we are postulating. 

797: 

798: After the three small planets merged to form a $6.3M_\oplus$ body, the

799: resulting planet was captured into the $3/1$ mean motion resonance

800: with Jupiter. Subsequently the eccentricities of both bodies increased

801: (we employed a rather low value of $\beta/\ep^2$ for this run). Both

802: planets migrated inward until the inner planet struck Jupiter, when

803: Jupiter was at $0.12$ AU, with an eccentricity of $0.4$.

804: 

805: In this run most of the mass in the disk actually accreted onto the

806: Jupiter mass planet. Part of the reason for this is that the smaller

807: planets were started at small radii, where the escape velocity from

808: the system exceeded the escape velocity from the Jupiter mass planet;

809: in that case we expect that most planetesimals will accrete onto

810: Jupiter rather than be ejected from the system. Only one body was

811: ejected. On the other hand, no bodies hit the star.

812: 

813: The latter result is representative of most of our runs; the fraction

814: of mass accreting onto the star is small, $\sim5\%$. The fraction

815: ejected varies with the initial semimajor axis and the mass of the

816: Jupiter mass body; both larger initial $\ap$ and larger $\mp$ produce

817: a larger fraction of ejected bodies (relative to bodies accreted onto

818: the massive planet). These rather low accretion fractions are in stark

819: contrast to those found by \cite{qh}. The difference appears to be

820: that we employ massive planetesimals, while their simulations employed

821: only test particles, which did not interact with each other. In our

822: simulations only the second most massive body (the first being

823: ``Jupiter'') remains for long in a resonance; this second most massive

824: body lords it over his smaller brethren, kicking them out of nearby

825: resonances they might like to occupy. This tends to prevent the

826: smaller objects from reaching the extremely high eccentricities

827: ($>0.9$) needed to strike the star.

828: 

829: In other runs the final state includes two planets in a mean motion

830: resonance. Since we start with such low planetesimal masses and

831: numbers, the mass ratio was always large. However, we expect that if

832: we allow larger terrestrial bodies to grow, that we may well find

833: final states with mass ratios nearer to unity. Finally, we note that

834: recent simulation of planetesimal migration show that two Jupiter-mass

835: object placed in a planetesimal disk will on some occasions migrate

836: toward each other \cite{hansen}. This could lead to resonance

837: capture followed by inward migration. Interactions between the massive

838: bodies and the planetesimal disk would likely tend to damp the

839: eccentricity of both bodies, but equation (\ref{both}) indicates that

840: as long as the outer body lost energy at a higher rate, the

841: eccentricities of both bodies would grow.

842: 

843: \section{DISCUSSION}

844: 

845: The mechanism we have proposed for the growth of eccentricity with

846: inward migration is essentially the same as that used to explain the

847: non-zero eccentricities of the inner Jovian satellites. It is well

848: understood and quite robust. It does not rely on the details of the

849: migration mechanism; in the case of the Jovian satellites the

850: migration is a result of the tidal bulge raised by Io on Jupiter; the

851: bulge exerts a torque on Io which transfers energy from Jupiter's spin

852: to the orbit of Io. Io in turn exerts, through a $2/1$ mean motion

853: resonance, a torque on Europa. In the satellite case the eccentricity

854: damping, produce by tidal flexing in both satellites as they oscillate

855: from peri- to apo-Jove, is very strong. This limits the eccentricity

856: to a value which, while small, is sufficient to dissipate enough

857: energy to power the volcanism on Io.

858: 

859: In previous sections we have examined resonant eccentricity growth in

860: the context of planetesimal migration, but it can work in the context

861: of migration due to tidal torques imposed by a gas disk as

862: well. Suppose that two Jupiter-mass bodies embedded in a gas disk are

863: locked in a mean motion resonance. Suppose that the gas between the

864: planets is removed, as numerical integrations indicate \cite{bryden}.

865: The gas inside the orbit of the inner planet will accrete onto the

866: star, possibly with some fraction being removed by a disk or stellar

867: wind. The planets are likely to follow the inner disk inward; if they

868: do not, the normal viscous spreading of the inner disk would move the

869: outer edge of the disk outward, until it experiences tidal torques

870: from the inner planet; this interaction would produce a back reaction

871: which would tend to damp the eccentricity of that planet; large

872: planetary eccentricities are unlikely to arise in that case.

873: 

874: However, it may be possible that the inner disk drains onto the star,

875: leaving the planets behind. This could occur, for example, if the

876: outer disk had a mass only slightly larger than that of the planets

877: \cite{npmk}. Then only the outer planet will experience significant

878: tidal torques, since the first order resonances of the inner planet

879: lie in the region between the planets that is depleted of gas.  Both

880: planets will then migrate inward, and eccentricity of the inner planet

881: will grow, since it does not experience much eccentricity

882: damping. This is exactly analogous to the planetesimal migration

883: described above, and the expressions we have given will describe the

884: growth and equilibrium values of the eccentricity once the appropriate

885: eccentricity damping rate for the outer planet is introduced (see,

886: e.g., Goldreich \& Tremaine 1980).

887: 

888: The tidal torque scenario also requires that the outer planet

889: have a mass sufficient to open a gap in the gas disk. If it does not,

890: then both bodies will experience tidal torques, which tend to damp

891: eccentricity rather strongly. An approximate criterion for gap

892: formation is \cite{lp86}

893: %

894: \be \label{simple_gap}%$

895: {m\over M_*}\gta 40\alpha (c_s/v_k)^2,

896: \ee %$

897: %

898: where $\alpha$ is the the \cite{ss} viscosity parameter, $c_s$ is the

899: sound speed in the gas disk, and $v_k$ is the Keplerian rotation

900: velocity. If the disk is ionized, then the Balbus-Hawley instability

901: \cite{bh91,hb91} is likely to produce a rather larger (effective)

902: $\alpha$, of order $0.5$. The planet must then have a mass of order

903: $200$ Jupiter masses ($m_J$) in order to open a gap.  At small orbital

904: radii $a\lta 0.1$ AU the disk will be ionized \cite{gammie}. This

905: suggests that if the capture into resonance occurs at very small

906: radii, or if the planets migrate to very small radii, the eccentricity

907: of both bodies will be damped.

908: 

909: However, at larger radii protoplanetary disks are believed to be

910: substantially neutral, so that they are not subject to the

911: Balbus-Hawley instability \cite{gammie}. If so, they will likely have

912: a small effective viscosity, and equation (\ref{simple_gap}) predicts

913: that gap opening will occur for small (subJovian) mass planets. In

914: terms of $\alpha$ currently favored values are in the range

915: $\alpha\approx10^{-4}$ to $10^{-2}$. The latter value yields a mass

916: for a gap clearing planet of about $2m_J$. Smaller values of $\alpha$

917: would yield smaller masses, but it is believed that in that case a

918: second criterion is relevant, namely $m/M_*>3(c_s/v_k)^2$

919: \cite{pl84}. At $5$ AU this yields $m\gta m_J$. Thus the eccentricity

920: of the inner planet will only be excited if both objects are of

921: roughly Jupiter mass, assuming the migration is driven by tidal

922: torques in a gas disk.

923: 

924: 

925: Another constraint is that the migration torque not exceed the

926: resonant torque. If it does, the resonance will be broken, and the

927: eccentricity of the outer body will drop. The outer body may then

928: migrate inward, perhaps to be caught into a stronger resonance. This

929: constraint is likely to be important in the capture phase,

930: particularly in a migration produced by tidal torques. In that

931: situation, both planets are likely to have very small

932: eccentricities. The tidal torque is given by  \cite{ward}

933: % 

934: \be \label{disk}%$

935: T_{disk}\approx \left({GM_*\mp\over 2a}\right){\mp\over M_*}

936: {M_{disk}\over M_*}\left({v_k\over c_s}\right)^3,

937: \ee %$

938: %

939: assuming that the outer planet does not open a gap. For a disk of mass

940: $M_{disk}=10^{-2}M_*$ and a planet at $5$ AU, this is about

941: $50(GM_*\mp/2a)$.  If the outer planet is massive enough to open a gap, the

942: torque is set by the viscosity in the gas disk,

943: %

944: \be %$

945: T_{gap}\approx\left({GM_*\mp\over 2a}\right)\alpha \left({c_s\over

946: v_k}\right)^2.

947: \ee %$

948: %

949: This is much smaller than the torque in the gapless case,

950: $T_{gap}\approx10^{-5}(\alpha/10^{-2})(GM_*\mp/2a)$.

951: The resonant torque is

952: %

953: \be %$

954: T_{res}\approx \left({GM_*\mp\over 2a}\right){m_1\over \mp}\left(

955: {ea_1\over a_1-\ap}\right)^q,

956: \ee %$

957: %

958: where $e$ is the larger of the eccentricities of the two planets. This

959: eccentricity is likely to be small; if we take $e\approx 0.01$ then

960: the resonant torque is $T_{res}\approx 10^{-2}(m_1/\mp)(GM_*\mp/ 2a)$

961: for the $2/1$ first order resonance; near the inner edge of the gap

962: (where $a/\Delta a\approx 10$) this will rise by about 10.

963: 

964: If both bodies are of roughly Jupiter mass, capture is possible into

965: first or second order resonances, that is, resonances with

966: $|j_1-\jp|=q$, where $q=1$ or $2$, since $T_{gap}$ is the

967: appropriate torque to use.  This case may arise in the scenario

968: mentioned above, where gas caught between two giant planets can leak

969: out over several hundred orbital periods. The tidal torques from the

970: gas inside the inner planet will then tend to push it outward, while

971: the gas outside the outer planet will tend to push it inward; capture

972: into the $2/1$ or possibly the $3/1$ mean motion resonance could then

973: occur.

974: 

975: If the outer body has a mass substantially smaller than $m_J$, it will

976: not open a gap in the gas disk. If it has a mass comparable to or

977: larger than $1M_\oplus$, the hydrodynamic drag it experiences will be

978: much smaller than the tidal torques. It will undergo rapid (Type I in

979: the notation of Ward) inward migration, easily passing through any

980: mean motion resonances (see equation \ref{disk}). According to Ward,

981: the time for this inward migration will be less than $10^5$ years. The

982: inward migration will not halt until the outer body enters the gap

983: produced by the inner, Jovian mass object. It will then experience a

984: torque similar to that felt by the inner, Jupiter mass body, and both

985: bodies will migrate inward without a substantial change in their

986: eccentricity.

987: 

988: Since the initial inward migration is so rapid, it seems unlikely that

989: an outer planet with initial $m_1\lta 10M_\oplus$ will be able to

990: accrete sufficient solid material to trigger the accretion of gas

991: before it enters the gap produced by the inner planet. Once it enters

992: the gap, the outer planet could grow by eating other, inward migrating

993: bodies, a la the scenario proposed by Ward (1997) for explaining the

994: very short period Jupiter mass objects. However, unlike Ward's case,

995: the outer planet cannot emerge far enough from the inner edge of the

996: outer disk that its $2/1$ resonance leaves the disk, slowing the

997: inward migration; the inner planet is in the way. Given the mismatch

998: between the tidal and resonant torques, it seems likely that the

999: smaller planet will be subsumed by the more massive body.

1000: 

1001: There may be ways to distinguish migration by tidal torques and

1002: migration by ejection of planetesimals. Planetesimal migration is

1003: likely to produce dynamically isolated (although possibly not

1004: unaccompanied) Jupiter mass bodies in small, eccentric

1005: orbits; as we have seen, the small mass inner body responsible for

1006: driving the eccentricity up to large values is often ingested into the

1007: star or the Jupiter mass object.

1008: 

1009: In those cases where the inner body survives the migration process, it

1010: should be possible to detect it with high precision radial velocity

1011: observations. In some cases the inner body may have a large mass,

1012: since planetesimal migration involving more than one Jupiter mass

1013: object sometimes produces a convergence of the semimajor axes of two

1014: bodies \cite{hansen}. This might produce systems like those recently

1015: discovered around GJ876 \cite{876}. Alternately, such a system could

1016: be the outcome of the migration of two Jupiter-mass planets in a gas

1017: disk. 

1018: 

1019: Resonant migration in a gas disk is  less

1020: likely to produce a single body in a moderately eccentric orbit than

1021: is migration in a planetesimal disk; in the former case

1022: both bodies are likely to be deep in resonance, and hence protected

1023: from close encounters and the subsequent carnage. If they do suffer

1024: close encounters, merger rather than ejection or accretion onto the

1025: central star is the likely result. In a planetesimal migration, the

1026: low mass of the inner planet combined with the frequency of close

1027: encounters with numerous smaller bodies tends to keep the amplitude of

1028: libration large. We have seen several cases in our numerical

1029: integrations where the inner planet collides with the Jupiter mass

1030: body, or with the central star.

1031: 

1032: There are several systems which have low mass planets in highly

1033: eccentric orbits, including HD108147 ($M\sin i=0.35M_J$, $a=0.098$ AU,

1034: $e=0.56$, and $K=37$m/s), HD83443 ($M\sin i=0.17M_J$, $a=0.174$ AU,

1035: $e=0.42$, and $K=14$m/s) (the system contains at least two planets,

1036: currently not dynamically linked), and HD16141 ($M\sin i=0.22M_J$,

1037: $a=0.351$ AU, $e=0.28$, and $K=10.8$m/s). The first system is

1038: particularly interesting as a test of the type of migration involved,

1039: assuming the eccentricity is produce by resonant migration. The

1040: putative resonant planet must have a mass less than about $1/3$ the

1041: mass of the detected planet, in order to escape detection (since the

1042: survey is clearly capable of finding planets with $K\approx 10$ m/s);

1043: this would give $M\sin i\approx0.1M_J$. For a typical inclination we

1044: would expect a mass of about of about $60M_\oplus$. Whether such a low

1045: mass object could open a gap in a gas disk is an interesting question.

1046: We note that the planet in HD108147 is near the radius at which the

1047: Balbus-Hawley instability is believed to operate. In a gas disk, a

1048: planet in this region would be subject to rapid eccentricity

1049: damping. As radial velocity surveys improve below the $10$m/s level,

1050: the discovery of even lower mass objects in eccentric orbits would

1051: indicate that some mechanism other than resonant migration in a gas

1052: disk was operating to produce the high eccentricities.

1053: 

1054: 

1055: The fact that our simulations show accretion of planetesimals onto the

1056: star suggests another way to distinguish the two scenarios. The

1057: planetesimal migration is inevitably accompanied by the accretion of

1058: $\sim5\%$ of the planetesimal disk mass onto the star; we see this even

1059: in simulations in which we halt the migration at large semimajor

1060: axis. Since the mass of the disk is of order $300-600M_\oplus$, this

1061: amounts to $\sim20M_\oplus$ of rocky material accreted onto the

1062: star. This material will include $5-7M_\oplus$ of iron, altering the

1063: apparent metallicity of the parent stars dramatically. Note that

1064: Jupiter contains only about $2M_\oplus$ of iron.

1065: 

1066: Moderate period (longer than 40 day period) systems produced by gas

1067: migration are unlikely to pollute their parent stars so dramatically;

1068: there is no reason to expect that such systems have a few hundred

1069: Earth masses of rocky material lying around. The parent stars could

1070: accrete metal rich Jupiter mass bodies after reaching the main

1071: sequence; but again there is no reason to expect that every moderate

1072: period system did so, when it is known that most stars that lack such

1073: companions did not \cite{mcahn}. We say this because planets with 40

1074: day or longer periods and moderate eccentricities ($\sim0.5$ or less)

1075: are dynamically uncoupled from very short period planets (four days or

1076: so), and so are unlikely to cause them to fall onto the star.

1077: 

1078: 

1079: \section{CONCLUSIONS}

1080: We have described how two resonant planets undergoing inward migration

1081: can reach eccentricities of order $0.7$. The eccentricity growth is

1082: largest when the inner planet is not subject to eccentricity

1083: damping. Such a situation may arise either in planetesimal migration,

1084: or in migration driven by tidal torques in which the inner gas disk

1085: has been removed by accretion or mass loss in a wind.  We have

1086: presented expressions for the equilibrium eccentricity, when it

1087: exists, and for the relation between $e$ and the initial and final

1088: semimajor axis of the resonant planets. 

1089: 

1090: We have presented numerical integrations showing that in some cases a

1091: planetesimal migration will produce a single Jupiter mass object with

1092: a large eccentricity; in other cases the Jupiter-mass object may be

1093: accompanied by a resonant object with a similar mass, or by a

1094: Neptune-mass companion. In the case of a Neptune-mass body the inner

1095: companion will have a very large eccentricity.

1096: 

1097: We have also described integrations involving $10$ or more roughly

1098: Earth mass bodies, together with a Jupiter mass planet; the latter is

1099: forced to migrate inward, with the migration process tending to damp

1100: its eccentricity. It typically captures one or more of the less

1101: massive bodies into mean motion resonance. Usually only the second

1102: most massive planet (which may not be the body that was originally

1103: second most massive behind ``Jupiter'') survives long in resonance. In

1104: fact, this second most massive body tends to accrete its smaller

1105: companions.

1106: 

1107: We have suggested two possible ways to tell the difference between

1108: systems produced by planetesimal or gas migration. First, if there

1109: exist a large number of systems having an eccentric planet with either

1110: no resonant companion, or with only very low mass (no-gap opening)

1111: companions, a finding requiring very high precision radial velocity

1112: measurements, it would strongly suggest that resonant migration by

1113: tidal torques was not responsible. Second, the planetesimal migration

1114: picture predicts that several Earth masses of iron will be accreted in

1115: the migration process, well after the central star has reached the

1116: main sequence. The tidal torque scenario is mute regarding this point;

1117: accretion of material after the gas disk vanishes, and closely

1118: correlated with the presence of Jupiter-mass planets, is not a natural

1119: feature.

1120: 

1121: 

1122: 

1123: \appendix

1124: 

1125: \section{Appendix}

1126: 

1127: To derive the expression for the variation of the eccentricities of

1128: two planets caught in a mean motion resonance when one or both are

1129: subject to dissipative forces, we examine the equations describing the

1130: evolution of energy and angular momentum, subject to the constraint

1131: that the planets are locked in a resonance \cite{lpc, gomes98}.

1132: 

1133: We start with the energy, which is given by

1134: %

1135: \be %$

1136: E=-{G\ms m_1\over 2a_1}-{G\ms \mp\over 2\ap}.

1137: \ee %$

1138: %

1139: We assume $a_1<\ap$. It will prove useful to employ the variables

1140: $\L_i\equiv\sqrt{GM_*a_i}$; then the resonance condition (\ref{lock})

1141: implies

1142: %

1143: \be \label{action}%$

1144: {d\ln \L_1\over dt}={d\ln {\cal L}_p\over dt}.

1145: \ee %$

1146: %

1147: This equation is accurate only in an average sense; on times shorter

1148: than the libration time of the resonance it is violated.

1149: 

1150: 

1151: We assume that some dissipative force removes both energy and angular

1152: momentum from the orbits;

1153: %

1154: \be %$

1155: {dE\over dt}=\left({dE_1\over dt}\right)_T+

1156: \left({d\Ep\over dt}\right)_T,

1157: \ee %$

1158: %

1159: where the subscript $T$ (for ``tides'') represents the effect of the

1160: non-conservative force. After some algebra, the energy evolution

1161: equation yields

1162: %

1163: \be %$

1164: \mp{d{\cal L}_p\over dt}+{\jp\over j_1}m_1{d\L_1\over dt}=

1165: \mp\left({d{\cal L}_p\over dt}\right)_T + 

1166: {\jp\over j_1}m_1\left({d\L_1\over dt}\right)_T .

1167: \ee %$

1168: %

1169: Combining this with equation (\ref{action}) we find

1170: %

1171: \be \label{appendix_energy}%$

1172: {d\ln \L_1\over dt}={d\ln {\cal L}_p\over dt}={1\over

1173: 1+\left(m_1/\mp\right)\left(\jp/j_1\right)^{2/3}}

1174: \left[\left({d\ln {\cal L}_p\over dt}\right)_T+

1175: {m_1\over \mp}\left({\jp\over j_1}\right)^{2/3}

1176: \left({d\ln \L_1\over dt}\right)_T\right] 

1177: \ee %$

1178: %

1179: 

1180: Next we examine the angular momentum

1181: %

1182: \be %$

1183: L=m_1\sqrt{G\ms a_1(1-e_1)^2}+\mp\sqrt{G\ms \ap(1-\ep)^2},

1184: \ee %$

1185: %

1186: which evolves according to

1187: %

1188: \be %$ 

1189: {dL\over dt}=\left({dL_1\over dt}\right)_T+\left({dLp\over

1190: dt}\right)_T.  \ee %$

1191: %

1192: Introducing the auxillary variables $Y_i=\sqrt{1-e_i^2}$ we find

1193: %

1194: \begin{eqnarray} \label{angular} %$

1195: \lefteqn{\left[{dY_1\over dt}-\left({dY_1\over dt}\right)_T\right]

1196: +{\mp\over m_1}\left({\jp\over j_1}\right)^{1/3}

1197: \left[{d\Yp\over dt}-\left({d\Yp\over dt}\right)_T\right]} \nonumber\\

1198: &&=Y_1\left[\left({d\ln \L_1\over dt}\right)_T -{d\ln \L_1\over dt}\right]

1199: +{\mp\over m_1}\left({\jp\over j_1}\right)^{1/3}

1200: \Yp\left[\left({d\ln {\cal L}_p\over dt}\right)_T -{d\ln {\cal L}_p\over dt}\right].

1201: \end{eqnarray} %$

1202: %

1203: Combining equations (\ref{appendix_energy}) and (\ref{angular})

1204: gives (Gomes 1998)

1205: %

1206: \begin{eqnarray} %$

1207: \lefteqn{\left[{dY_1\over dt}-\left({dY_1\over dt}\right)_T\right]

1208: +{\mp\over m_1}\left({\jp\over j_1}\right)^{1/3}

1209: \left[{d\Yp\over dt}-\left({d\Yp\over dt}\right)_T\right]} \nonumber \\

1210: &&=

1211: \left\{

1212: \left({d\ln \L_1\over dt}\right)_T-\left({d\ln {\cal L}_p\over dt}\right)_T

1213: \right\}

1214: \left[Y_1-{\jp\over j_1}\Yp\right]

1215: \Bigg/

1216: \left[

1217: 1+{m_1\over \mp}\left({\jp\over j_1}\right)^{2/3}

1218: \right].

1219: \end{eqnarray} %$

1220: %

1221: 

1222: Writing this in terms of $e_i$ and $a_i$, we find

1223: %

1224: \begin{eqnarray} %$

1225: \lefteqn{

1226: {e_1\over \sqrt{1-e_1^2}}\left[{de_1\over dt}

1227: -\left({de_1\over dt}\right)_T\right] 

1228: +{\mp\over m_1}\left({\jp\over j_1}\right)^{1/3}

1229: {\ep\over \sqrt{1-\ep^2}}\left[{d\ep\over

1230: dt}-\left({d\ep\over dt}\right)_T\right] } \nonumber \\

1231: &&=\left\{

1232: \left({d\ln a_1\over dt}\right)_T-\left({d\ln \ap\over dt}\right)_T

1233: \right\}

1234: \left[

1235: {\jp\over j_1}\sqrt{1-\ep^2}-\sqrt{1-e_1^2}

1236: \right]

1237: \Bigg/2

1238: \left[

1239: 1+{m_1\over \mp}\left({\jp\over j_1}\right)^{2/3}

1240: \right].\label{both}

1241: \end{eqnarray} %$

1242: %

1243: 

1244: 

1245: \acknowledgements Support for this work was provided by NSERC of

1246: Canada.

1247: 

1248: \clearpage

1249: 

1250: \begin{thebibliography}{}

1251: 

1252: \bibitem[Balbus \& Hawley 1991]{bh91}

1253: Balbus, S.\ A.\ \& Hawley, J.\ F.\ 1991a, \apj, 376, 214 

1254: 

1255: \bibitem[Bryden et al. 2000]{bryden}

1256: Bryden, G., R{\'o}{\.z}yczka, M., Lin, D.\ N.\ C., \& Bodenheimer, P.\

1257: 2000, \apj, 540, 1091  

1258: 

1259: \bibitem[Ford et al. 2001]{ford}

1260: Ford, E. B., Havlickova, M. \& Rasio, F. A. 2001, to appear in Icarus (astro-ph/0010178)

1261: 

1262: \bibitem[Gammie 1996 ]{gammie}

1263: Gammie, C.\ F.\ 1996, \apj, 457, 355 

1264: 

1265: \bibitem[Goldreich 1965]{goldreich}

1266: Goldreich, P. 1965, \mnras, 130, 159

1267: 

1268: \bibitem[Goldreich \& Tremaine 1980]{GT}

1269: Goldreich, P.\ \& Tremaine, S.\ 1980, \apj, 241, 425 

1270: 

1271: \bibitem[Gomes 1998]{gomes98}

1272: Gomes, R. 1998, \aj, 116, 997

1273: 

1274: \bibitem[Gonzalez et al. 2001]{gonzalez} 

1275: Gonzalez, G., Laws, C., Tyagi, S., \& Reddy, B.\ E.\ 2001, \aj, 121, 432 

1276: 

1277: \bibitem[Hansen et al. 2001]{hansen}

1278: Hansen, B., Murray, N. \& Holman, M. 2001, \apj, submitted.

1279: 

1280: \bibitem[Hawley \& Balbus 1991]{hb91} 

1281: Hawley, J.\ F.\ \& Balbus, S.\ A.\ 1991b, \apj, 376, 223 

1282: 

1283: \bibitem[Holman \& Murray 1996]{hm}

1284: Holman, M. \& Murray, N. 1996, \aj, 112 1278 

1285: 

1286: \bibitem[Laughlin 2000 ]{laughlin}

1287: Laughlin, G.\ 2000, \apj, 545, 1064 

1288: 

1289: \bibitem[Lin et al. 1996]{lbr}

1290: Lin, D.\ N.\ C., Bodenheimer, P., \& Richardson, D.\ C.\ 1996, \nat,

1291: 380, 606  

1292: 

1293: \bibitem[Lin \& Papaloizou 1986]{lp86}

1294: Lin, D. N. C. \& Papaloizou, J. C. B. 1986, \apj, 307, 395

1295: 

1296: \bibitem[Lissauer et al. 1984]{lpc}

1297: Lissauer, J. J., Peale, S. J. \& Cuzzi, J. N. 1984, {\em Icarus}, 58, 159

1298: 

1299: \bibitem[Marcy et al. 2001]{876}

1300: Marcy, G. W., Butler, R. P., Fischer, D., Vogt, S. S., Lissauer,

1301: J. J. \& Rivera, E. J., 2001, \apj, submitted.

1302: 

1303: \bibitem[Murray et al. 2001]{murray}

1304: Murray, N., Chaboyer, B. \& Noyes, R. W. 2001, \apj, submitted

1305: 

1306: \bibitem[Murray et al. 2001]{mcahn} Murray, N., Chaboyer, B., Arras,

1307: P., Hansen, B. \& Noyes, R. W. 2001, \apj, to appear (preprint

1308: astro-ph/0011530)

1309: 

1310: \bibitem[Murray et al. 1998]{mhht}

1311: Murray, N., Hansen, B., Holman, M. \& Tremaine, S. 1998 {Science},

1312: 279, 69

1313: 

1314: \bibitem[Nelson et al. 2000]{npmk} 

1315: Nelson, R.\ P., Papaloizou, J.\ C.\ B., Masset, F. \& Kley, W.\ 

1316: 2000, \mnras, 318, 18 

1317: 

1318: \bibitem[\"Opik 1976]{opik}

1319: \"Opik, E.J. 1976 {\em Interplanetary Encounters} (Elsevier Scientific,

1320: Amsterdam)

1321: 

1322: \bibitem[Papaloizou et al. 2001]{PNM}

1323: Papaloizou, J.C.B., Nelson, R.P. \& Masset, F. 2001, \aap, 366, 263

1324: 

1325: \bibitem[Papaloizou \& Lin 1984]{pl84}

1326: Papaloizou, J.\ \& Lin, D.\ N.\ C.\ 1984, \apj, 285, 818 

1327: 

1328: \bibitem[Quillen \& Holman 2000]{qh} 

1329: Quillen, A.\ C.\ \&  Holman, M.\ 2000, \aj, 119, 397 

1330: 

1331: \bibitem[Santos et al. 2000]{santos}

1332: Santos, N.\ C., Israelian, G., \& Mayor, M.\ 2000, \aap, 363, 228 

1333: 

1334: \bibitem[Shakura \& Sunyaev 1973]{ss}

1335: Shakura, N. I. \& Sunyaev, R. A. 1973, \aap, 24, 373

1336: 

1337: \bibitem[Ward 1997]{ward}

1338: Ward, W.\ R.\ 1997, \apjl, 482,  L211 

1339: 

1340: \bibitem[Wisdom 1980]{wisdom} Wisdom, J.\ 1980, \aj, 85, 1122 

1341: 

1342: 

1343: \end{thebibliography}

1344: 

1345: %

1346: 

1347: 

1348: \clearpage

1349: 

1350: \begin{figure}

1351: \plotone{fig1.eps}

1352: \caption[The $E-L$ plane]{The $E-L$ plane for a planet with

1353: $e_p=0.1$. The curved solid line is the maximum angular momentum

1354: possible for a body of the given energy; it corresponds to a circular

1355: orbit ($e=0$). The dotted line corresponds to a planetesimal orbit that just

1356: grazes the orbit of the Jupiter mass body. The solid line labeled

1357: $J=-3/2$ illustrates the minimum value of the Jacoby parameter that

1358: must be reached by the asteroid in order to be lifted from an orbit

1359: with $a<a_p$ to an orbit with $a>a_p$. In this figure we employ units

1360: in which $GM_*=\ap=1$.

1361: %}

1362: \label{Fig_EL}}

1363: \end{figure}

1364: 

1365: 

1366: \begin{figure}

1367: \plottwo{fig2a.eps}{fig2b.eps}

1368: \caption[$dJ/dE$]{a) The time averaged change $dJ/dE$ for a

1369: planetesimal with initial eccentricity $e=0.05$, as a function of its

1370: initial $a$ (dashed curve), and the corresponding $\beta$ (solid

1371: curve). The planet is assumed to have a mass equal to that of Jupiter

1372: ($318M_\oplus$), and $e_p=0.05$. The solid vertical line marked $\mu^{2/7}$

1373: marks the region where first order mean motion resonances overlap,

1374: producing large scale chaos. The solid vertical line near $a=0.71$

1375: marks the location of the $5/3$ mean motion resonance. Note that

1376: $dJ/dE\approx 0.03$ near the $\mu^{2/7}$ region. The horizontal line

1377: corresponds to $\gamma$ for an inner planet of mass $10M_\oplus$

1378: trapped in the $4/1$ mean motion resonance.  b)

1379: Same as in (a), except that $e_p=e=0.5$. The horizontal line

1380: corresponds to an inner planet with a mass of $40M_\oplus$.

1381: %}

1382: \label{Fig_dJdE}}

1383: \end{figure}

1384: 

1385: \begin{figure}

1386: \plottwo{fig3a.eps}{fig3b.eps}

1387: \caption[$e$ vs. $t$]{The evolution of a system of 2 bodies, one

1388: having a mass of $40M_\oplus$, placed inside the orbit of the second,

1389: a Jupiter mass object that is forced to migrate inward. The inner body

1390: is placed just inside the $3/1$ mean motion resonance. Panel (a) shows

1391: the semimajor axis of both bodies as a function of time, while (b)

1392: shows the eccentricity; the light line (higher $e$) corresponds to the

1393: inner, lighter planet.

1394: \label{Fig_basic}}

1395: \end{figure}

1396: 

1397: 

1398: \begin{figure}

1399: \plottwo{fig4a.eps}{fig4d.eps}

1400: \caption[crash]{The evolution of a system of 5 bodies with masses

1401: randomly distributed between $0.3$ and $10M_\oplus$, placed inside the

1402: orbit of a Jupiter mass object that is forced to migrate inward. Three

1403: of the small planets merged with each other to create a $6.3M_\oplus$

1404: mass planet. One of the small planets merged with the Jupiter mass

1405: planet, while the fifth small planet was ejected. Panel (a)

1406: shows the semimajor axis of one of the low mass planets that merged to

1407: form the $6.3M_\oplus$ mass body; the mass

1408: increases with time due to collisions with other bodies, seen as jumps

1409: in $a$. This body was caught into resonance with the Jupiter mass

1410: planet at $t=9\times10^6$ years. (b) The eccentricity of the Jupiter

1411: mass body as a function of time. It damps slowly up until the time it

1412: captures the smaller mass body in (a), then rises rapidly. The value

1413: of $\beta$ used in this run assumed that $e=0.05$, which is not

1414: appropriate for the final $\ep\approx 0.4$.

1415: \label{Fig_crash}}

1416: \end{figure}

1417: 

1418: 

1419: \end{document}

1420: 

1421: 

1422: 

1423: 

1424: