1: \documentclass[preprint]{aastex}
2: %\documentclass[manuscript,letter,epsfig,amsmath]{aastex}
3: %\usepackage{emulateapj5}
4:
5: \newcommand{\etal}{{et al.\ }}
6: \newcommand{\mas}{\mu\mathrm{as}}
7: \newcommand{\be}{\begin{equation}}
8: \newcommand{\ee}{\end{equation}}
9: \newcommand{\bea}{\begin{eqnarray}}
10: \newcommand{\eea}{\end{eqnarray}}
11:
12: \shorttitle{Hot Jupiters and the Roche limit}
13: \shortauthors{Ford \& Rasio}
14:
15: \begin{document}
16: %
17: \title{On the Relation Between Hot Jupiters and the Roche Limit}
18: \author{Eric B.\ Ford$^1$, Frederic A.\ Rasio$^2$}
19: \affil{$^1$Astronomy Department, UC Berkeley, 601 Campbell Hall,
20: Berkeley, CA 94709}
21: \email{eford@astro.berkeley.edu}
22: \affil{$^2$Physics and Astronomy, Northwestern University, 2145
23: Sheridan Rd,
24: Evanston, IL 60208}
25: \email{rasio@northwestern.edu}
26: %
27: %
28: \begin{abstract}
29: %
30: Many of the known extrasolar planets are ``hot Jupiters,''
31: giant planets with orbital periods of just a few days.
32: %
33: We use the observed distribution of hot Jupiters to constrain the
34: location of its inner edge in the mass--period diagram. If we
35: assume a slope corresponding to the classical Roche
36: limit, then we find that the edge corresponds to a separation close
37: to {\it twice\/} the Roche
38: limit, as expected if the planets started on highly
39: eccentric orbits that were later circularized. In contrast,
40: any migration scenario would predict an inner edge right at the
41: Roche limit, which applies to planets approaching on nearly circular
42: orbits.
43: However, the current sample of
44: hot Jupiters is not sufficient to provide a precise constraint
45: simultaneously
46: on both the location and slope of the inner edge.
47: %
48: %
49: \end{abstract}
50:
51: \keywords{planetary systems: formation --- methods: statistical}
52:
53: \section{Introduction}
54: %
55: Early discoveries of hot Jupiters hinted at
56: a pile-up near a 3-day period, but recent transit surveys and more
57: sensitive
58: radial velocity observations have discovered planets with even shorter
59: periods. The data now suggest that the inner limit for hot
60: Jupiters is not defined by an orbital period, but rather by a tidal
61: limit, which depends on both the separation and the planet-star
62: mass ratio (Fig.\ 1). This would arise naturally if the inner
63: edge were related to the Roche limit, the critical distance within
64: which a
65: planet would start losing mass (Faber et al.\ 2005). The Roche limit
66: separation,
67: $a_R$, is given by $ R_P = 0.462\, a_R \,\mu^{1/3}$,
68: where $R_P$ is the radius of the planet, and $\mu=m/M_*$ is the
69: planet-star
70: mass ratio.
71:
72: The many formation scenarios proposed for hot Jupiters can be divided
73: into two broad
74: categories. The first involves slow migration on quasi-circular
75: orbits, perhaps due to
76: interaction with a
77: gaseous disk or planetesimal scattering (Murray et al.\ 1998; Trilling \etal~1998).
78: This would result in an inner edge precisely at the Roche limit.
79: The second category invokes tidal circularization of highly eccentric
80: orbits with very small pericenter distances, following
81: planet-planet scattering (Rasio \& Ford 1996; Weidenschilling \&
82: Marzari 1996;
83: Ford et al.\ 2001; Papaloizou \& Terquem 2001; Marzari \&
84: Weidenschilling
85: 2002), secular perturbations from a wide binary companion (Holman et
86: al.\ 1997;
87: Wu \& Murray 2003), or tidal capture of free-floating planets
88: (Gaudi 2003). These would result in a limiting separation of {\it
89: twice\/} the Roche limit, assuming that circularization can take
90: place without
91: significant mass loss from the planet\footnote{This is very easy to show:
92: consider a planet on an initially
93: eccentric orbit, with initial eccentricity $e$ and
94: pericenter distance $r_p$. Circularizing this orbit under ideal
95: conditions leads to dissipation of energy but conservation of mass and
96: angular momentum. Simply equating the angular momentum of the initial
97: and final orbits gives a final circularized radius
98: $a = r_p (1+e) \simeq 2 r_p$ for $e \simeq 1$.} (Faber et al.\ 2005;
99: Gu et al.\ 2003; Rasio et al.\ 1996).
100:
101: \section{Statistical Analysis}
102:
103: To constrain rigorously the
104: distribution of hot Jupiters, we adopt a Bayesian
105: framework, where the model parameters are
106: treated as random variables to be constrained by the actual
107: observations. To perform a Bayesian analysis it is
108: necessary to specify both the likelihood (the probability of making a
109: certain observation given a particular set of model parameters) and
110: the prior (the {\em a priori\/} probability distribution for the
111: model parameters). Let us denote the model parameters by $\theta$ and
112: the data by $d$, so that their joint probability
113: distribution function (PDF) is given by
114: $
115: p(d, \theta) = p(\theta) p(d | \theta) = p(d) p(\theta | d).
116: $
117: Note how the joint PDF is expanded in two ways,
118: both expressed as the product of a marginalized PDF
119: and a conditional PDF. The prior is
120: given by $p(\theta)$ and the likelihood by $p(d|\theta)$, while $p(d)$
121: is the {\em a priori\/} probability for observing the
122: values actually measured and $p(\theta|d)$ is the PDF
123: of primary interest: the {\em a posteriori\/} PDF
124: for the model parameters conditioned on the actual
125: observations. The probability of the
126: observations $p(d)$ can be
127: obtained by marginalizing over the joint PDF and again
128: expanding the joint density as the product of the prior and the
129: likelihood. This leads to Bayes' theorem, the primary tool for
130: Bayesian inference,
131: \be
132: p(\theta | d ) = \frac{p(d|\theta)p(\theta)}{p(d)} = \frac{p(d|\theta)
133: p(\theta)}{\int \, d\theta p(d|\theta)p(\theta)}
134: \ee
135: Often the model parameters contain a quantity of
136: particular interest (the location of the inner cutoff for hot
137: Jupiters in our analysis) plus other ``nuisance parameters,'' which are
138: necessary to describe the observations (e.g., the fraction
139: of stars with hot Jupiters in our analysis). Since Bayes' theorem
140: provides a real PDF for the model parameters, we
141: can simply marginalize over the nuisance parameters to calculate a
142: marginalized posterior PDF, which will be our basis
143: for making inferences about the location of the inner cutoff for hot
144: Jupiters.
145:
146: \subsubsection{1-d Model}
147:
148: We start by presenting a simple 1-d model for the
149: distribution of hot Jupiters. The primary question we wish to
150: address is the location of the inner edge of the
151: distribution relative to the Roche
152: limit. Therefore, we define $x \equiv a / a_R$, where $a$ is the
153: semimajor axis of the planet and $a_R$ is the Roche limit. We assume
154: that the actual distribution of $x$ for hot Jupiters is given
155: by a truncated power law,
156: %
157: \be
158: p(x | \gamma, x_l, x_u) dx = x^\gamma \left(\frac{dx}{x}\right),
159: \quad x_l < x < x_u,
160: \ee
161: %
162: and zero elsewhere. Here $\gamma$ is the power-law index and $x_l$
163: and $x_u$ are the lower and upper limits for $x$. The lower limit,
164: $x_l$, is the model parameter of primary interest, while $\gamma$ and
165: $x_u$ are nuisance parameters. Therefore, our results are contained in
166: the marginalized posterior PDF for $x_l$.
167:
168: For simplicity we restrict our analysis to the subset of
169: known extrasolar planets discovered by complete radial-velocity surveys,
170: extremely unlikely to contain any false positives. To
171: obtain such a sample, we impose two constraints: $P\le P_{\max}$, where
172: $P_{\max}$ is a maximum orbital period, and $K\ge K_{\min}$, where
173: $K_{\min}$ is a minimum velocity semi-amplitude. We use
174: $K_{\min}=30$m/s, following Cumming (2004). We typically set $P_
175: {\max}=30\,$d, even though
176: radial-velocity surveys are likely to be complete for even longer
177: periods (provided $K\ge K_{\min}$). This minimizes the chance of
178: introducing biases due to survey incompleteness or possible structure
179: in the
180: observed distribution at larger periods. By
181: considering only planets with orbital parameters such that radial-
182: velocity
183: surveys are very nearly complete, our analysis does not
184: depend on the velocities of stars for which no planet has been
185: detected. Note that our criteria for including a planet may
186: introduce a bias depending on the actual mass-period distribution.
187: We will address this with a 2-d model below.
188: Note that, in this paper, we exclude any planet
189: discovered via techniques other than radial velocities (e.g.,
190: transits), even if subsequent radial-velocity observations were
191: obtained to confirm the planet.
192:
193: Initially, we make several simplifying assumptions to allow for a
194: simple analytic
195: treatment. We assume uniform priors
196: for each of the model parameters, $p(\gamma) \sim
197: U(\gamma_{\min},\gamma_{\max})$ and $p(x_l,x_u) \sim \mathrm{const}$,
198: provided $x_{ll} < x_l < x_u < x_{uu}$ and zero otherwise. The lower
199: and upper limits are chosen
200: to be sufficiently far removed from regions of high likelihood that
201: these choices do not affect our results.
202: %
203: We assume that the orbital period ($P$), velocity semi-amplitude
204: ($K$), semi-major axis ($a$), stellar mass ($M_*$), and planet mass
205: ($m \sin i$) are known exactly based on the observations.
206:
207: We begin by assuming that $\sin i=1$ (orbital plane seen nearly edge-
208: on) for all systems
209: and that all planets have the same radius, $R_P$.
210: With these
211: assumptions, the posterior probability distribution is
212: %
213: \be
214: p(x_l, x_u, \gamma | x_1, ... x_n ) \sim \gamma^n (x_u^\gamma - x_l^
215: \gamma)^{-n}
216: \prod_{j=1}^n x_j^{\gamma-1},
217: \ee
218: %
219: provided that
220: $x_{ll} < x_l \le x_{(1)} \le x_{(n)} \le x_u < x_{uu}$ and
221: $\gamma_{\min} < \gamma < \gamma_{\max}$. Here
222: $n$ is the number of planets included
223: in the analysis, $x_{(1)}$ is the smallest value of $x$ among the
224: planets used in the analysis, and $x_{(n)}$ is the largest
225: value. The normalization can be obtained by
226: integrating over all allowed values of $x_l$, $x_u$, and $\gamma$.
227: %
228: We show the marginal posterior distributions after
229: integrating over the nuisance parameters, $x_u$ and $\gamma$, in Fig.~2
230: (dotted line), assuming $R_P = 1.2\,R_J$. The distribution has a sharp
231: cutoff at $x_{(1)}$ and a tail to lower values reflecting the chance
232: that $x_l < x_{(1)}$ due to the finite sample size.
233:
234: Next, we adopt an isotropic distribution of inclinations
235: ($\cos i\sim~U[-1,1]$), but we use the measured value
236: for radial-velocity planets where the orbital
237: inclination has been determined via transits. We show the marginal
238: posterior distribution for $x_l$ in Fig.~2 (solid line).
239: The sharp cutoff at $x_{(1)}$ is replaced with a more gradual tail,
240: reflecting the chance that $\sin i<1$ for planets with the smallest
241: values of $x$.
242:
243: Now consider the consequences of allowing for a distribution of
244: planetary radii. For transiting planets we use a normal distribution
245: based on the published radius value and uncertainty. For
246: non-transiting planets, we assume
247: a normal distribution with standard deviation $\sigma_{R_P}$. We
248: show the resulting
249: marginalized posterior PDFs in Fig.~2. Allowing for a
250: significant dispersion broadens the posterior distribution for $x_l$
251: and results in a slight shift to smaller values.
252: We have also explored the effects of varying the model parameter
253: $P_{\max}$ from $8\,$d to $60\,$d. We find that this does not
254: produce any discernible difference in the posterior PDF for $x_l$.
255:
256: Our results are sensitive to the choice of mean radius for the
257: non-transiting planets. In Fig.~3 (upper) we show the posterior
258: PDFs for various mean radii, assuming $\sigma_{R_P}
259: = 0.1\,R_J$. Since few planets have a known inclination, there is
260: a nearly perfect degeneracy between $R_P$ and $x_l$. Even when
261: we include transiting planets, this degeneracy remains near perfect,
262: i.e., $p(x_l | R_p, x_1, ..., x_n ) \simeq p(x_l \cdot \frac{R'_P}
263: {R_P} | R'_P, x_1, ..., x_n)$.
264: However, it remains extremely
265: unlikely that $x_l\simeq1$ for any reasonable planetary
266: radius.
267:
268: \subsubsection{2-d Model}
269:
270: We can improve our analysis by more properly considering the
271: joint PDF in orbital period and planet-star mass ratio, which we
272: write as
273: %
274: \be
275: p(P, \mu | \alpha, \beta, P_{\min}, P_{\max}, \mu_{\min}, \mu_{\max},
276: c ) \sim c P^\alpha \mu^\beta \frac{dP}{P} \frac{d\mu}{\mu},
277: \ee
278: %
279: provided $\mu_{\min} < \mu < \mu_{\max}$, $P < P_{\max}$, and $a(P,
280: M_*) \ge x_l \cdot a_R(R_P, \mu)$. Here $\alpha$ and $\beta$ are new
281: power-law indices. Again our
282: results are not sensitive to the nuisance
283: parameters, $\mu_{\min}$, $\mu_{\max}$, and $P_{\max}$. For
284: definiteness, we fix their values at $P_{\max} = 30\,$d, $\mu_{\min}
285: = 3.3\times10^{-5}$, and $\mu_{\max} = 0.01$. We take priors uniform in
286: $\tan^{-1}(\alpha)$ and $\tan^{-1}(\beta)$, as the density
287: $U[-\pi/2,\pi/2]$ corresponds to uniform prior density for the slope
288: of the power-law distribution on a log-log plot. We take a prior
289: uniform in $\log c$, as is standard for scale parameters. Our
290: calculation of the likelihood is similar to that of Tabachnik \&
291: Tremaine (2002), except we replaced their inner boundary of $P\ge
292: P_{\min}$ with our boundary $a(P, M_*) \ge x_l a_R(R_P,\mu)$. The
293: necessary integrals can be performed analytically provided we
294: approximate $1+\mu\simeq1$. For convenience we used
295: {\tt Maple} to obtain an analytic expression
296: for the integral of the likelihood over $P$ and $\mu$. The remaining
297: integrals over $\alpha$, $\beta$, and $c$ were performed numerically
298: over a grid with $\sim10^{10}$ points.
299:
300: By considering the joint mass-period PDF, we are
301: able to account for the bias previously introduced by imposing
302: $K\ge K_{\max}$. Since we are considering only planets with
303: orbital parameters such that radial-velocity surveys are very nearly
304: complete, our results depend only on the number of surveyed stars for
305: which no planet has been discovered, and not the observed velocities
306: of these stars. For the total number of stars in radial-velocity
307: surveys that are complete for $K\ge K_{\min}$ and
308: $P\le P_{\max}$ we estimate $N_* = 2000$.
309: We show the resulting marginalized posterior PDF for
310: $x_l$ in Fig.\ 3 (lower).
311:
312: \subsubsection{The Shape of the Inner Cutoff}
313:
314: The above results clearly demonstrate that the present
315: observations strongly favor an inner cutoff at the ideal
316: circularization distance rather than at the Roche limit, {\em
317: assuming that
318: the inner edge follows the slope of the Roche limit\/}. We have also
319: performed
320: calculations treating this slope
321: as an unknown model parameter. Unfortunately, the present
322: observations are not
323: sufficient to constrain this parameter empirically, and the resulting
324: marginalized posterior PDF for $x_l$ still allows, but no longer
325: exclusively favors, $x_l \simeq 2$.
326:
327: We could gain additional leverage by including planets in short-period
328: orbits down to lower $K_{\min}$. Unfortunately, the incompleteness of
329: the radial-velocity surveys is likely to be significant for $K_{\min}
330: < 20\,$m/s. Since the number, quality, and spacing of observations
331: varies widely among the stars in these surveys, it would be
332: necessary to calculate the probability for detecting planets as a
333: function of orbital period and velocity semi-amplitude for each
334: star.
335: Additionally, the planet mass-radius relation, which is extremely
336: flat for planets near $1\,M_J$, becomes important for much
337: lower-mass planets. Therefore, we have not attempted to extend our
338: analysis to planets for which radial-velocity surveys are not
339: yet complete. We expect that the recent improvements in measurement
340: precision will eventually extend their completeness to smaller
341: $K_{\min}$.
342:
343: We have begun a preliminary investigation of the constraints obtained
344: by adding information from the OGLE transit survey (Udalski 2002).
345: Due to both signal-to-noise and aliasing issues, the OGLE survey does
346: not provide a complete sample of short-period planets for a
347: significant fraction of parameter space. Therefore, it is necessary
348: to calculate the probability of detecting planets with various orbital
349: periods and radii. We estimate these probabilities by taking the actual
350: observation times and uncertainties for
351: the 62 transit candidates from the 2002 OGLE observations of the
352: Carina field and applying the detection criteria from Pont et al.\ (2005).
353: While the transiting planets are consistent with our above findings,
354: they do not provide sufficient additional information to constrain the
355: slope of the inner edge. We look forward to both radial-velocity
356: and transit surveys detecting additional lower-mass objects in
357: short-period orbits, so that we may eventually constrain the slope of
358: the cutoff empirically.
359:
360: \section{Discussion}
361:
362: The current distribution of hot Jupiters shows a cutoff that is a
363: function of orbital period and
364: planet mass. Under the assumption that the slope of this cutoff
365: follows the Roche limit,
366: our Bayesian analysis solidly rejects the hypothesis
367: that the cutoff occurs inside or at the present Roche limit. This is
368: in constrast to
369: what would be expected if these planets had slowly migrated inwards
370: on quasi-circular orbits and with radii close to the presently measured
371: values around $1.2\,R_{J}$. If confirmed by future analyses of a more
372: extensive data set, this result would be highly significant, as it
373: would eliminate
374: a broad class of popular migration scenarios for the formation of hot
375: Jupiters.
376:
377: Instead, our analysis shows that this cutoff
378: occurs at a distance nearly twice that of the Roche limit, as expected
379: if the planets had been circularized from a highly eccentric orbit.
380: These findings suggest that hot Jupiters may have formed via
381: planet-planet scattering (Rasio \& Ford 1996), tidal capture of
382: free floating planets (Gaudi 2003), or secular perturbations from a
383: highly inclined binary companion (Holman et al.\ 1997). Regardless
384: of the
385: exact mechanism,
386: our model would require that the hot Jupiters all started on highly
387: eccentric
388: orbits and survived the strong tidal dissipation needed to
389: circularize their
390: orbits. A few caveats are worth mentioning.
391: Our study addresses the statistical properties
392: of the population of hot-Jupiters and does not attempt to advance the
393: state of knowledge of any specific planet.
394: In particular, we adopt
395: average properties of an assumed distribution that is analogous
396: to---and derived from---the presently known distribution of hot Jupiters,
397: but we do not consider or solve for the specific properties of any
398: individual extrasolar planet.
399: Moreover, strongly non-random or non-gaussian effects would be poorly modeled with the
400: technique developed here.
401:
402: An alternative explanation is that the planets migrated inwards at
403: an early time and arrived at their Roche limit on a quasi-circular orbit
404: when their radii were still $\ge 2\,R_{J}$ (Burrows et al.\ 2000). The
405: dissipation process causing the migration must then have stopped
406: immediately
407: afterwards to avoid further decay of the orbit as the planets
408: continued to
409: cool and contract. We find this scenario unattractive, especially since
410: there is no natural explanation for the factor of 2 in this case.
411:
412: Yet another alternative is that short-period giant planets are destroyed
413: by some process {\em before\/} they reach the Roche limit. HST
414: observations of HD 209458 indicate absorption by matter presently
415: beyond the Roche lobe of the planet and have been interpreted as
416: evidence
417: for a wind leaving the planet and powered by stellar irradiation
418: (Vidal-Madjar et
419: al.\ 2003, 2004). Further theoretical work will help determine under
420: what conditions these processes can cause significant mass loss
421: (e.g., Hubbard et al.\ 2005) and whether complete destruction could
422: occur rather
423: suddenly when the orbital radius decreases below $\sim 2 a_{R}$.
424:
425: Future planet discoveries will either tighten the
426: constraints on the model parameters or provide evidence for the
427: existence of planets definitely closer than twice the Roche limit.
428: Additionally, future discoveries of
429: transiting hot-Jupiters around young stars could help discriminate
430: between
431: the above alternatives.
432: Moreover, new detections of lower-mass planets with very short
433: periods could help better
434: constrain the shape of the inner cutoff as a function of mass. In the
435: future, an improved statistical analysis could also include such low-
436: mass planets,
437: where surveys are not yet complete.
438:
439:
440: \acknowledgments{We thank E.\ Chiang, T.\ Loredo, N.\ Murray,
441: R.\ Murray-Clay, J.\ Papaloizou, F.\ Pont, and an anonymous
442: referee for helpful comments.
443: %
444: This research was supported by NSF grants AST-0206182 and AST-0507727
445: at Northwestern University and by a Miller Research Fellowship to EBF
446: at UC Berkeley.
447: }
448: %
449: \begin{references}
450: %
451: \reference{} Burrows, A., et al. 2000, ApJ, 534, 97
452:
453: \reference{} Cumming, A. 2004, MNRAS, 354, 1165
454:
455: \reference{} Faber, J.A., Rasio, F.A. \& Willems, B. 2005, Icarus,
456: 175, 248
457:
458: \reference{} Ford, E.B., Havlickova, M., \& Rasio, F.A. 2001, Icarus,
459: 150, 303
460:
461: \reference{} Gaudi, S. 2003, astro-ph/0307280.
462:
463: \reference{} Gu, P.-G., Lin, D.N.C. \& Bodenheimer P.H. 2003, ApJ,
464: 558, 509
465:
466: \reference{} Holman, M., Touma, T. \& Tremaine, S. 1997, Nature, 386,
467: 254
468:
469: \reference{} Hubbard, W.B., Hattori, M.F., Burrows, A., Hubeny, I., \&
470: Sudarsky, D. 2005, astro-ph/0508591
471:
472: \reference{} Marzari, F. \& Weidenschilling, S.J. 2002, Icarus 156, 670
473:
474: \reference{} Murray, N., Hansen, B., Holman, M., \& Tremaine, S. 1998,
475: Science, 279, 69
476:
477: \reference{} Papaloizou, J.C.B. \& Terquem, C. 2001, MNRAS, 325, 221
478:
479: \reference{} Pont et al.\ 2005, A\&A 438, 1123
480:
481: \reference{} Rasio, F.A. \& Ford, E.B. 1996, Science, 274, 954
482:
483: \reference{} Rasio, F.A., Tout, C.A., Lubow, S.H., \& Livio, M.
484: 1996, ApJ, 470, 1187
485:
486: \reference{} Tabachnik, S. \& Tremaine, S. 2002, MNRAS 335, 151
487:
488: \reference{} Trilling et al.\ 1998, ApJ, 500, 428
489:
490: \reference{} Udalski et al.\ 2002, Acta Astron., 52, 1
491:
492: \reference{} Vidal-Madjar, A., Lecavelier des Etangs, A., Desert, J.-
493: M., Ballester, G.E., Ferlet, R., Hebrand, G., Mayor, M. 2003, Nature,
494: 422, 143
495:
496: \reference{} Vidal-Madjar, A. et al. 2004, ApJL 604, 69
497:
498: \reference{} Weidenschilling, S.J., \& Marzari, F. 1996, Nature,
499: 384, 619
500:
501: \reference{} Wu, Y., \& Murray, N. 2003, ApJ, 589, 605
502:
503: \end{references}
504:
505: %
506: \begin{figure}[ht]
507: \plotone{f1.eps}
508: \caption[fig1]{
509: %
510: %\hbox{\plotone{f1.eps} } \noindent Fig.\ 1:
511: %
512: Minimum mass ratio vs orbital period for the current observed
513: sample. Planets discovered by radial velocity
514: surveys are shown as triangles with arrows indicating $1-\sigma$
515: uncertainties in mass due to unknown inclination. The magenta squares
516: were discovered by radial velocity observations and have inclinations
517: and radii measured via transits. The blue squares show planets
518: discovered by transit searches. The green lines show the minimum mass
519: corresponding to various velocity semi-amplitudes and roughly indicate
520: where radial velocity surveys are nearly complete ($\ge30\,$m/s), have
521: significant sensitivity ($\ge10\,$m/s), and are only beginning to detect
522: planets ($\ge3\,$m/s). The two red lines show the location of the
523: Roche limit ($a_R$) and the ideal circularization radius ($a_{\rm
524: circ}$) for a planet with a radius $R_P = 1.2\,R_J$. The red
525: lines do not apply to the lowest mass planets that likely have a
526: radius significantly less than $1.2\,R_J$ given their
527: different internal structure. \\
528: %
529: \label{Fig1}}
530: \end{figure}
531:
532: %\noindent
533:
534:
535: \begin{figure}[ht]
536: \plotone{f2.eps}
537: \caption[fig2]{
538: %
539: %\hbox{\plotone{f2.eps} } \noindent Fig.\ 2:
540: %
541: Marginalized posterior probability distribution for $x_l$, the
542: lower cutoff for the ratio of a planet's semimajor axis
543: to the Roche limit. Here we show multiple posterior distributions for
544: various simplified models. The dotted curve assumes $R_P = 1.2\,R_J$
545: and $\sin i=1$ for all planets, while the remaining lines assume the
546: observed inclinations and radii for transiting planets and an
547: isotropic distribution of inclinations and mean radius $R_P =
548: 1.2\,R_J$ for the remaining planets. The widths of the distributions in
549: radii are $0.0 R_J$ (dotted and solid curves), $0.05 R_J$ (long dashes),
550: $0.1 R_J$ dots-dashes), $0.2 R_J$ (short dashes). \\
551: %
552: \label{Fig2}}
553: \end{figure}
554:
555: %\noindent
556:
557: %
558: \begin{figure}[ht]
559: \plotone{f3.eps}
560: \caption[fig3]{
561: %
562: %\hbox{\plotone{f3.eps} } \noindent Fig.\ 3:
563: %
564: Posterior distributions from our 1-d model (top) and full 2-d model
565: (bottom) for
566: various mean radii: $\left<R_P\right> = 1.0
567: R_J$ (long dashes), $1.1 R_J$ (dotted), $1.2 R_J$ (solid), $1.3 R_J$
568: (dotted dashed), and $1.4 R_J$ (short dashes), all assuming $\sigma_
569: {R_P}
570: = 0.1 R_J$. \\
571: %
572: \label{Fig3}}
573: \end{figure}
574:
575: %\noindent
576:
577: \end{document}
578: