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

2: %\documentclass[preprint,onecolumn]{emulateapj}

3: 

4: \newcommand{\etal}{{et al.\ }}

5: \newcommand{\be}{\begin{equation}}

6: \newcommand{\ee}{\end{equation}}

7: \newcommand{\bea}{\begin{eqnarray}}

8: \newcommand{\eea}{\end{eqnarray}}

9: \newcommand{\go}{\mathrel{\raise.3ex\hbox{$>$}\mkern-14mu\lower0.6ex\hbox{$\sim$}}}

10: \newcommand{\lo}{\mathrel{\raise.3ex\hbox{$<$}\mkern-14mu\lower0.6ex\hbox{$\sim$}}}

11: 

12: \shorttitle{Eccentricities of Planets from Photometry}

13: \shortauthors{Ford et al.~}

14: %\slugcomment{submitted to ApJ}

15: 

16: \begin{document}

17: %

18: \title{Characterizing the Orbital Eccentricities of Transiting Extrasolar Planets with Photometric Observations}

19: 

20: \author{Eric B.\ Ford\altaffilmark{1,2,3}, Samuel N.\ Quinn\altaffilmark{2}, Dimitri Veras\altaffilmark{3,4,5}}

21: 

22: \altaffiltext{1}{Hubble Fellow}

23: %

24: \altaffiltext{2}{Harvard-Smithsonian Center for Astrophysics, Mail Stop 51, 60 Garden Street, Cambridge, MA 02138, USA} %; eford@cfa.harvard.edu}

25: %

26: \altaffiltext{3}{Department of Astronomy, University of Florida, 211 Bryant Space Science Center, P.O. Box 112055, Gainesville, FL 32611-2055, USA}

27: %

28: \altaffiltext{4}{JILA, Campus Box 440, University of Colorado, Boulder, CO 80309, USA}

29: %

30: \altaffiltext{5}{Department of Astrophysical and Planetary Sciences, University of Colorado, Boulder, CO 80309, USA}

31: %

32: 

33: \begin{abstract}

34: %

35: The discovery of over 200 extrasolar planets with the radial velocity

36: (RV) technique has revealed that many giant planets have large

37: eccentricities, in striking contrast with most

38: of the planets in the solar system and prior theories of planet

39: formation.  The realization that many giant planets have large

40: eccentricities raises a fundamental question: ``Do terrestrial-size

41: planets of other stars typically have significantly eccentric orbits

42: or nearly circular orbits like the Earth?''  

43: Here, we demonstrate that photometric observations of transiting

44: planets could be used to characterize the orbital eccentricities for

45: individual transiting planets, as well the eccentricity distribution

46: for various populations of transiting planets (e.g., those with a

47: certain range of orbital periods or physical sizes).  Such

48: characterizations can provide valuable constraints on theories for the

49: excitation of eccentricities and tidal dissipation.  We outline the

50: future prospects of the technique given the exciting prospects for

51: future transit searches, such as those to be carried out by the CoRoT

52: and Kepler missions.

53: %

54: \end{abstract}

55: %

56: \keywords{planetary systems --- planetary systems: formation --- planets 

57: and satellites: general  --- techniques: photometric --- methods: statistical

58:  --- celestial mechanics}

59: %

60: 

61: \section{Introduction}

62: \label{sec_intro}

63: %

64: 

65: Theorists have proposed numerous mechanisms that could excite orbital

66: eccentricities.  Some of these mechanisms are expected to affect all

67: planets independent of their mass (e.g., perturbations by binary

68: companions, passing stars, or stellar jets; e.g., Holman et al.\ 1997;

69: Laughlin \& Adams 1998; Ford et al.\ 2000; Zakamska \& Tremaine 2004;

70: Namouni 2007), while the efficiency of other mechanisms would vary

71: with planet mass (e.g., planet-disk or planet-planet interactions;

72: e.g., Artymowicz 1992; Goldreich \& Sari 2003; Chatterjee et al.\

73: 2007).  If the mechanism(s) exciting eccentricities of the known giant

74: planets also affect terrestrial planets, then Earth-mass planets on

75: nearly circular orbits could be quite rare.  On the other hand, if

76: large eccentricities are common only in systems with massive giant

77: planets and/or very massive disks, then there may be an abundance of

78: planetary systems with terrestrial planets on low eccentricity orbits

79: (Beer et al.\ 2004).  Thus, understanding the eccentricity

80: distribution of terrestrial planets could provide empirical

81: constraints for planet formation theories (e.g., Ford \& Rasio 2007)

82: and shed light on the processes that determined the eccentricity

83: evolution in our solar system.  Since the discovery of transiting

84: giant planets with eccentric orbits, authors have begun to consider

85: the implications of eccentricities for transiting planets (e.g.,

86: Barnes 2007; Burke 2008).

87:  

88: %

89: The CoRoT and Kepler space missions are expected to detect many

90: transiting planets and measure their sizes and orbital periods,

91: including some in or near the ``habitable zone'' (e.g., Kasting et al.\

92: 1993).  The Kepler mission aims to determine

93: the frequency of Earth-like planets, and study how the frequency and

94: properties of planets correlates with the properties of their host

95: stars (Basri et al.\ 2005).  

96: %

97: Since a significant eccentricity would cause the stellar flux incident

98: on the planet's surface to vary, a planet's eccentricity affects its

99: climate (i.e., equilibrium temperature, amplitude of seasonal

100: variability) and potentially its habitability (Williams et al.\ 2002;

101: Gaidos \& Williams 2004).  Our method could be applied to these

102: planets to determine the frequency of terrestrial planets that could

103: also have Earth-like climates, and thus influence the design of future

104: space missions that will attempt to detect and characterize nearly

105: Earth-like planets (e.g., Space Interferometry Mission-PlanetQuest)

106: and search them for signs of life (e.g., Terrestrial Planet Finder).

107: 

108: 

109: In \S\ref{SecDuration}, we show how the duration of a transit is

110: affected by a planet's orbital eccentricity.  We outline how to

111: interpret the transit duration for transiting planets with both low

112: signal-to-noise light curves (\S\ref{SecLoSN}) and high

113: signal-to-noise (\S\ref{SecHiSN}).  We compare the magnitude of the

114: effect on the transit duration to the expected precision of

115: eccentricity constraints based on Kepler photometric data

116: (\S\ref{SecStatErr}) and also the typical accuracy of stellar

117: parameters (\S\ref{SecStarErr}).

118: %

119: We demonstrate that Kepler observations could be used to characterize

120: the orbital eccentricities of terrestrial planets.  We describe

121: statistical approaches for analyzing the distribution of transit

122: durations of a population of transiting planets in \S\ref{SecStats}.

123: In \S\ref{SecRV}, we discuss the role of radial velocity observations

124: for constraining eccentricities of Earth-like transiting planets.  In

125: \S\ref{SecDiscuss}, we conclude with a discussion of how the results

126: could contribute to understanding the formation and evolution of

127: terrestrial planets and address fundamental questions, such as ``What

128: is the frequency of terrestrial planets that have Earth-like

129: eccentricities?'' and ``What is the frequency of terrestrial planets

130: that pass through the habitable zone?''

131: 

132: \section{Light Curve of an Eccentric Transiting Planet}

133: \label{SecLightCurve}

134: 

135: %\subsection{Relation of Transit Duration to Orbital Eccentricity}

136: \label{SecDuration}

137: %

138: The total transit duration is defined as the time interval between the

139: first and fourth points of contact (Fig. \ref{FigGeo}).  In the

140: approximation that the mean planet-star separation ($a$) is much

141: greater than both the stellar radius ($R_{\star}$) and the planetary

142: radius ($R_p$), the transit duration is much less than the orbital

143: period ($P$), and thus the planet-star separation during the transit

144: ($d_t$) is nearly constant.  Using these approximations, the total

145: transit duration ($t_D$) is given by

146: %

147: \be 

148: \frac{t_D}{P} \simeq \frac{R_{\star}}{\pi a\sqrt{1-e^2}}

149: \sqrt{\left(1+r\right)^2-b^2} \left(\frac{d_t}{a}\right), 

150: \ee 

151: %

152: where $e$ is

153: the orbital eccentricity, $r\equiv R_p/R_{\star}$ is the ratio of

154: the planet radius to stellar radius, $b\equiv d_t \cos i / R_{\star}$

155: is the impact parameter, and $i$ is the orbital inclination

156: measured relative to the plane of the sky (Tingley \& Sackett 2005).

157: %

158: The planet-star separation at the time of transit is given by $d_t = a

159: ( 1 - e \cos E_t ) =a (1-e^2)/(1+e \cos T_t)=a (1-e^2)/(1+e \cos

160: \omega)$, where $T_t$ and $E_t$ are the true and eccentric anomalies

161: at the time of transit, and $\omega$ is the argument of periastron

162: measured relative to the line of sight.  (Note that this differs from

163: the convention for radial velocity determinations.)  Thus, the ratio

164: of the actual transit duration to the transit duration for the same

165: planet on a circular orbit (assuming other transit parameters such as

166: size and impact parameter are held fixed) is $(1+e

167: \cos\omega)/\sqrt(1-e^2)$.  We have

168: performed numerical integrations to verify that this approximation is

169: accurate to better than $\simeq0.1\%$, for eccentricities as high as

170: 0.95 (Fig.\ \ref{fig_tauvsomega}).

171: 

172: 

173: We define the variable

174: %

175: %\begin{equation}

176: $

177: \tau_b \equiv d_t/(a\sqrt{1-e^2})

178: $

179: %\end{equation}

180: %

181: to be the ratio of the transit duration of one planet to the transit duration for an

182: identical planet with the same orbital period and impact parameter,

183: but on a circular orbit.

184: %

185: In general, transiting planets could have significant eccentricities,

186: in which case $\tau_b$ will deviate from unity (e.g., $\tau_b\simeq1.8$ for HAT-P2b; Bakos et al.\

187: 2007).  For planets on low eccentricity orbits,

188: $\left|\tau_b-1\right|\simeq O(e)$.  For large eccentricities, $\tau_b$

189: can range from zero to several (Fig.\ \ref{fig_tauvsomega}).

190: %

191: For the case of a central transit of an Earth-like planet

192: orbiting a solar-mass star on a circular orbit,

193: %

194: \be

195: % t_D \simeq 13 hours \left(\frac{P}{\mathrm{yr}}\right) \left(\frac{R_{\star}}{R_\odot}\right) \left(\frac{AU}{a}\right) \left(\frac{\left(1+r\right)^2-b^2}{1-e^2}\right)^{\frac{1}{2}} \left(\frac{d_t}{a}\right),

196: %

197: t_D \simeq 13 \mathrm{hours} \left(\frac{P}{\mathrm{yr}}\right)^{1/3} \left(\frac{R_{\star}}{R_\odot}\right) \left(\frac{M_{\odot}}{M_{\star}}\right)^{1/3} \left(1+\mu\right)^{-1/3} \left(\left(1+r\right)^2-b^2\right)^{1/2}  \tau_b ,

198: \ee

199: %

200: where $M_{\star}$ is the stellar mass and $\mu\equiv m_p/M_{\star}$ is

201: the planet-star mass ratio.  The planet-star mass ratio ($\mu\equiv

202: m_p/M_{\star}$) is typically negligible for planetary companions.  (If

203: it is large, then it could be measured by obtaining RV observations.)

204: By assuming that the pericenter direction is randomly oriented relative

205: to the line of sight, we can compute the probability distribution for

206: $\tau_b$ as a function of eccentricity (Fig.\ \ref{FigProbTau}).

207: 

208: 

209: 

210: \subsection{Analysis of Transiting Planet Light Curves}

211: \label{SecAnalysis}

212: %

213: Here we outline the basics of how photometric light curves can be used

214: to constrain the eccentricities of transiting planets.  We discuss two

215: limiting regimes: 1) low signal-to-noise (S/N) light curves that

216: provide no constraint on the impact parameter, and 2) high S/N light

217: curves that measure the impact parameter.

218: 

219: \subsubsection{Impact Parameter Not Measured (Low S/N)}

220: \label{SecLoSN}

221: %

222: For any transit detection, photometric observations alone directly

223: measure the orbital period ($P$), the total transit duration ($t_D$),

224: and the depth of the transit.  For the purposes of providing analytic

225: estimates, we neglect the effects of limb darkening, so that the

226: transit depth is given by the planet-star area ratio ($r^2$).  

227: %

228: For some faint stars and/or small planets, it may be difficult to

229: measure additional light curve parameters such as the ingress duration

230: and the impact parameter.  For these cases, the total transit

231: duration, $t_D$, can be compared to $t_{D,o}$, the transit

232: duration expected for the same planet with the same orbital period

233: ($P$) and mean stellar density ($\rho_{\star}$), but assuming a

234: circular orbit and central transit ($b=0$).  Even for transiting

235: planets with low S/N light curves, photometric observations can measure the ratio

236: %

237: \begin{equation}

238: \tau_o \equiv \frac{t_D}{t_{D,o}} = \left(\frac{d_t}{a\sqrt{1-e^2}}\right)  \left(\frac{\sqrt{\left(1+r\right)^2-b^2}}{1+r}\right) \simeq \left(\frac{t_D}{13\mathrm{hours}}\right) \left(\frac{\mathrm{yr}}{P}\right)^{1/3} \left(\frac{\rho_{\star}}{\rho_{\odot}}\right)^{1/3}  \frac{\left(1+\mu\right)^{1/3}}{\left(1+r\right)}.

239: \label{EqnTauo}

240: \end{equation}

241: %

242: Since $\mu=m_p/M_{\star}\ll 1$ for planetary mass companions, a value

243: of $\tau_o$ significantly greater than unity can only arise for an

244: eccentric planet, but a value of $\tau_o$ less than unity could be due

245: to either a non-central transit ($b>0$) or a non-zero eccentricity

246: ($e>0$).  Thus, it is not possible to measure the

247: eccentricity for an individual planet with only low S/N photometry.

248: However, it is still possible to characterize the distribution of

249: eccentricities for a population of planets based on the observed

250: distribution of $\tau_o$ (see \S\ref{SecStatsKs}).  Since the above analysis can be applied

251: to relatively faint stars with low S/N transit light curves, we expect

252: that Kepler will discover many planets that can be included in 

253: statistical analyses of $\tau_o$ (Basri et al.\ 2005).  

254: %

255: We note the actual analysis should consider the effects of limb

256: darkening, which cause the depth of a high-latitude transit relative

257: to be greater than a than an equatorial transit even for the same

258: star-planet area ratio.  While limb darkening precludes simple

259: analytic expressions, such effects can be incorporated into Monte

260: Carlo simulations.

261: 

262: 

263: 

264: In

265: \S\ref{SecStats}, we will show that Kepler observations could

266: constrain the {\em distribution of eccentricities} and reject

267: plausible eccentricity distributions for terrestrial planets.

268: %

269: The precision of the eccentricity constraints will depend on the

270: precision and accuracy of the measurements of $\tau_o$.  In

271: \S\ref{SecStatErr} we show that the the orbital period, transit depth,

272: and transit duration can be measured with sufficient precision that

273: the accuracy for the measured $\hat{\tau}_o$ will typically be limited

274: by the deviation of the inferred stellar density

275: ($\hat{\rho}_{\star}$) from the actual stellar density

276: ($\rho_{\star}$), i.e., $\hat{\tau}_o\simeq \tau_o

277: (\hat{\rho}_{\star}/\rho_{\star})^{1/3}$.  The mean stellar density

278: could be estimated using spectroscopy and stellar modeling

279: (\S\ref{SecStarErr}), resulting in a typical accuracy for

280: $\hat{\tau}_o$ of $\sim 5-15\%$ (e.g., Ford et al.\ 1999; Fischer \& Valenti 2005; Takeda et al.\ 2007).  The spread of transit

281: durations due to the unknown inclination will have a negligible

282: effect on the ability of low S/N transit light curves to constrain the

283: eccentricity distribution.

284: 

285: 

286: 

287: \subsubsection{Impact Parameter Measured (High S/N)}

288: \label{SecHiSN}

289: %

290: For high-quality light curves, photometric observations can measure the time of each point of contact and determine both the

291: total transit duration ($t_D$) and $t_F$ (Fig.\ \ref{FigGeo}).  If we

292: neglect limb darkening, then 

293: % the duration of the nearly flat bottom of the transit light curve will be given by

294: %

295: %\be

296: %\frac{t_F}{P} = \frac{R_{\star}}{\pi a\sqrt{1-e^2}} \sqrt{\left(1-r\right)^2-b^2} \left(\frac{d_t}{a}\right),

297: %\ee

298: %

299: % Then 

300: the impact parameter can be determined by the photometric observables,

301: %

302: %\be

303: $

304: b^2 = \left[\left(1-r\right)^2-\gamma^2\left(1+r\right)^2\right]/\left(1-\gamma^2\right),

305: $

306: %\ee

307: %

308: where we define $\gamma^2\equiv t_F^2/t_D^2$ (Seager \& Mallen-Ornelas

309: 2003).  

310: 

311: In practice, the transit shape will be affected by limb darkening

312: (Fig.\ \ref{FigLimbDarkening}) and we must fit for the limb darkening

313: parameters to determine the impact parameter (and planet-star radius

314: ratio).  Previous experience analyzing transit photometry of giant

315: planets has found that the limb darkening parameters can be accurately

316: estimated with stellar models and the uncertainty in limb darkening

317: parameters typically introduces relatively modest additional

318: uncertainties and have only small correlations with the other model

319: parameters of interest (e.g., Brown et al.\ 2001; Holman et al.\ 2006;

320: Burke et al.\ 2007; Knutson et al.\ 2007; Winn et al.\ 2007; Torres et

321: al.\ 2008).  

322: %

323: In order to verify that the uncertainty in limb darkening coefficients

324: will introduce only modest additional uncertainties, we have generated

325: simulated Kepler V-band light curves for terrestrial planets

326: transiting a $V=12$ magnitude solar-like star, including a quadratic

327: limb darkening model (Mandell \& Agol 2002; Claret 2000).  We then

328: calculate $\Delta\chi^2$, varying the transit time, transit duration,

329: the planet size, impact parameter, and quadratic limb darkening

330: coefficients (see Fig.\ \ref{FigContours}).  While these

331: are some modest correlations between the transit duration and limb

332: darkening parameters, these correlations are a much smaller effect

333: than the uncertainty due to the correlation between the transit

334: duration and the impact parameter.  Therefore, we conclude that the

335: uncertainty in impact parameter will be the limiting factor and we can

336: neglect the effects of limb darkening when estimating the power of

337: transit durations for constraining the eccentricities of transiting

338: planets (see \S\ref{SecStatsKs}).

339: %

340: %ADD FIG HERE

341: %

342: Once the host star and transit time are known from Kepler

343: observations, a future transit can be observed using ground-based

344: observatories at multiple wavelengths.  By combining the inferred

345: stellar properties with limb darkening profiles from stellar models,

346: one can determine the impact parameter from even relatively low

347: precision data (e.g., Jha et al.\ 2000).

348: Infrared observations would be

349: particularly useful, as they minimize the effects of limb darkening

350: and aid in the measurement of the impact parameter.  Since the impact

351: parameter is the same at all wavelengths, ground-based transit light

352: curves could effectively eliminate the need to fit for this parameter

353: when analyzing the Kepler photometry.  Here, we assume that limb-darkening

354: parameters can be well constrained by some combination of stellar modeling

355: and external observations.  

356: %

357: %We note that the limb darkening due to a small planet is likely to be somewhat simpler than for a giant planet, since the blocked patch of the stellar surface is smaller and more uniform, allowing for a more precise determination of limb darkening parameters (Mandell \& Agol 2002).

358: 

359: For systems with high S/N photometry, the total transit duration,

360: $t_D$, can be compared to $t_{D,b}$, the transit duration expected for

361: the same planet and star with the same observed orbital period, impact

362: parameter ($b$), and mean stellar density, but

363: assuming a circular orbit. Then, 

364: Kepler can measure the ratio

365: %

366: \begin{equation}

367: \tau_b = \frac{t_D}{t_{D,b}} \simeq \left(\frac{t_D}{13\mathrm{hours}}\right) \left(\frac{\mathrm{yr}}{P}\right)^{1/3}  \left(\frac{1-\gamma^2}{r}\right)^{1/2} \left(\frac{\rho_{\star}}{\rho_{\odot}}\right)^{1/3}  \left(1+\mu\right)^{1/3}.

368: \label{EqnTauB}

369: \end{equation}

370: %

371: %

372: The fact that the planet is transiting the star places a constraint on a

373: combination of true anomaly and the direction of pericenter.  Therefore, 

374: the actual value of $\tau_b$ depends only on the eccentricity and

375: direction of pericenter, and $\tau_b$ can be measured using a combination

376: of transit photometry and stellar parameters (e.g.,  from spectroscopy and stellar modeling).

377: %

378: Using high S/N photometry, we can accurately measure the first

379: three terms on the right hand side of Eqn. \ref{EqnTauB}.  The last term containing the planet-star

380: mass ratio ($\mu$) is negligible for planetary

381: companions.

382: Therefore, the accuracy of the measured $\tau_b$ will typically be

383: limited by the uncertainty in the cube root of the mean stellar density

384: ($\sigma_{\rho_{\star}^{1/3}}$).

385: We can detect a non-zero eccentricity for 

386: an individual planet when

387: $e>\sigma_{\rho_{\star}^{1/3}}/\rho_{\star}^{1/3}\simeq

388: 0.05-0.15$  (\S\ref{SecStarErr}).  This is quite significant given the mean eccentricity 

389: of planets discovered by RV surveys is $\simeq 0.3$ (ignoring planets

390: with $P<10$ days that may have been influenced by tidal circularization; Butler et al.\ 2006).

391: 

392: 

393: \subsection{Expected Precision of Kepler Measurements}

394: \label{SecStatErr}

395: %

396: Here we present estimates of the timing precision possible with Kepler

397: observations, assuming a flux measurement precision of

398: $\sigma_{ph}\simeq$400ppm during a one minute integration on a V=12

399: star and uncertainties that scale with the square root of the photon

400: count (Basri et al.\ 2005).  For the sake of deriving approximate

401: analytic expressions, we assume uncorrelated Gaussian uncertainties

402: and ignore complications due to limb darkening in this paper

403: 

404: To determine the precision of eccentricity constraints, we must

405: estimate the precision of the constraints from transit photometry.

406: The fractional transit depth ($r^2$ in the absense of limb darkening)

407: can be measured with a precision, $\sigma_{r^2}\simeq \sigma_{ph} /

408: \sqrt{t_F \Lambda N_{tr}}$, where $\Lambda$ is the rate of photometric

409: measurements with precision $\sigma_{ph}$ and $N_{tr}\simeq 4

410: (\mathrm{yr}/P)$ is the number of transits observed during the four

411: year mission lifetime.  Note that in this approximation the precision

412: does not depend on the integration time, provided that it is

413: significantly less than the ingress/egress time.  In practice, an

414: increased rate of measurements is valuable for constraining any

415: brightness variations across the stellar disk (e.g., limb darkening,

416: star spots, plage).

417: 

418: The time of

419: each ingress/egress ($t_{in/out}$) can be measured with a precision

420: $\sigma_{t_{in/out}} \simeq \sigma_{ph} \sqrt{\Delta t_{in/out}/\Lambda}

421: \left(R_{\star}/R_p\right)^{2}$, where $\Delta t_{in/out}$ is the duration of ingress/egress (Ford \& Gaudi 2006).

422: %

423: Therefore, the sidereal orbital period can be measured with a

424: precision of $\sigma_P \simeq \sigma_{t_{in/out}}

425: /\sqrt{2\left(N_{tr}-1\right)}$ and the duration of the transit can be

426: measured with a precision of $\sigma_{t_D} = \sigma_{t_{in/out}}

427: \sqrt{2/N_{tr}}$.  Once there are at least two transits, the

428: uncertainty in $t_D$ will dominate.

429: 

430: The duration of ingress ($\Delta t_{in}$) or egress($\Delta t_{out}$) is given by

431: $\Delta t_{in/out}\equiv \epsilon t_D$, where $\epsilon

432: \left(1-\epsilon\right) =

433: r/\left[\left(1+r\right)^2-b^2\right]$.  For non-grazing

434: transits, $\epsilon\ll1$, so the ingress/egress duration is

435: %

436: %\be 

437: %\frac{t_{in/out}}{P} \simeq

438: %\frac{R_p}{\pi a\sqrt{1-e^2}} \left(\frac{d_t}{a}\right) / \sqrt{\left(1+r\right)^2-b^2}.

439: %\ee

440: %

441: %Again, focusing on the case of an Earth-like planet,

442: %

443: %\be

444: %t_{in/out} \simeq 7.1 minutes \left(\frac{P}{\mathrm{yr}}\right) \left(\frac{R_p}{R_\oplus}\right) \left(\left(1+r\right)^2-b^2\right)^{\frac{-1}{2}}\left(1-e^2\right)^{\frac{-1}{2}} \left(\frac{d_t}{a}\right), 

445: %\ee

446: %

447: \be

448: \Delta t_{in} \simeq \Delta t_{out} \simeq 7.1 \mathrm{minutes} \left(\frac{R_p}{R_\oplus}\right) \left(\frac{P}{\mathrm{yr}}\right)^{\frac{1}{3}} \left(\left(1+r\right)^2-b^2\right)^{\frac{-1}{2}} \left(\frac{M_{\odot}}{M_{\star}}\right)^{\frac{1}{3}} \left(1+\mu\right)^{\frac{-1}{3}} \tau_b.

449: % \left(1-e^2\right)^{\frac{-1}{2}} \left(\frac{d_t}{a}\right) ,

450: \ee

451: %

452: %

453: Inserting fiducial values for a target star with an apparent magnitude of $V$, 

454: % we expect a time precision for each single ingress/egress

455: %

456: %\be

457: %\sigma_{t_{in/out}} \simeq 13.8 min \left(\frac{P}{\mathrm{yr}}\right)^{1/2} \left(\frac{a}{AU}\right)^{-1/2}  \left(\frac{R_p}{R_\oplus}\right)^{-3/2}  \left(\frac{R_{\star}}{R_\odot}\right)^2 \left(\frac{d_t}{a}\right)  \left(1-e^2\right)^{-1/4} \left(\left(1+r\right)^2-b^2\right)^{-1/4} 10^{\left(V-12\right)/5}.

458: %\ee

459: %

460: %\be

461: %\sigma_{t_{in/out}} \simeq 13.8 \mathrm{min} \left(\frac{P}{\mathrm{yr}}\right)^{1/6} \left(\frac{R_p}{R_\oplus}\right)^{-3/2}  \left(\frac{R_{\star}}{R_\odot}\right)^2 \left(\frac{M_{\odot}}{M_{\star}}\right)^{1/6} \left(\frac{d_t}{a}\right) \frac{\left(1+\mu\right)^{-1/6} 10^{\left(V-12\right)/5}}{  \left(1-e^2\right)^{1/4} \left(\left(1+r\right)^2-b^2\right)^{1/4} }.

462: %\ee

463: %

464: % so the fractional measurement

465: we expect a fractional precision for the mean transit duration of an individual planet to be 

466: %

467: \be

468: \frac{\sigma_{t_D}}{t_D} \simeq 0.013 \left(\frac{P}{\mathrm{yr}}\right)^{\frac{1}{3}} \left(\frac{R_p}{R_\oplus}\right)^{\frac{-3}{2}}  \left(\frac{R_{\star}}{R_\odot}\right)  \left(\frac{M_{\star}}{M_{\odot}}\right)^{\frac{1}{6}} \left(1+\mu\right)^{\frac{1}{6}} \tau_b^{\frac{-1}{2}} \left(\left(1+r\right)^2-b^2\right)^{\frac{-3}{4}} 10^{\left(V-12\right)/5}.

469: \ee

470: %

471: %

472: For a Jupiter-sized planet with a similar orbit and host star, the

473: expected fiducial fractional uncertainty for Kepler observations

474: decreases to $\simeq3\times10^{-4}$.  Ground-based

475: observatories have achieved a precision of

476: $\sigma_{t_D}/t_D\simeq1.5\%$ for short-period giant planets (e.g.,

477: Holman et al.\ 2007).

478: 

479: 

480: For modest eccentricities, Kepler's expected measurement

481: precision will place a lower limit on the uncertainty of a given

482: planet's eccentricity ($\sigma_e$) of order $\simeq \sigma_{t_D}/t_D$.

483: Even for Earth-sized planets around V=14 stars,

484: Kepler is expected to achieve the photometric precision necessary to

485: measure eccentricities as low as $\simeq 0.03$.  

486: 

487: For the sake of completeness, we also estimate one additional

488: timescale. For an eccentric orbit, there is a slight change in $d_t$

489: between ingress and egress that results in a difference between the

490: ingress and egress durations,

491: %

492: \be

493: % t_{in}-t_{out} \simeq 4s \left(\frac{P}{\mathrm{yr}}\right) \left(\frac{R_p}{R_\oplus}\right) \left(\frac{R_{\star}}{R_\odot}\right) \left(\frac{AU}{a}\right)^2 \left(\frac{e \sin T_t}{\left(1-b^2\right) \left(1+e\cos T_t\right) \left(1-e^2\right)^{3/2} }\right),

494: %

495: \Delta t_{in}-\Delta t_{out} \simeq 4 \mathrm{sec} \left(\frac{P}{\mathrm{yr}}\right)^{-1/3} \left(\frac{R_p}{R_\oplus}\right) \left(\frac{R_{\star}}{R_\odot}\right) \left(\frac{M_{\odot}}{M_{\star}}\right)^{2/3} \left(\frac{\left(1+\mu\right)^{-2/3}  e \sin T_t}{\left(1-b^2\right)  \left(1+e\cos T_t\right) \left(1-e^2\right)^{3/2} }\right),

496: \ee

497: %

498: where $T_t$ is the true anomaly at the time of mid-transit and we have

499: neglected the second order effect of change in the impact parameter.

500: For a Jupiter-sized planet with a 4 day orbit around a solar-sized star,

501: the difference in ingress and egress durations increases to 3.4

502: minutes.  While Kepler and/or high-precision follow-up observations

503: might be able to measure such effects for a relatively small number of

504: bright stars, we expect other methods will typically provide more

505: powerful constraints on the eccentricity.

506: 

507: \subsection{Uncertainties in Stellar Mass and Radius}

508: \label{SecStarErr}

509: %

510: For many transiting planets photometric observations will be so

511: precise that uncertainties in the stellar properties will limit the

512: accuracy of $\hat{\tau}_o$ or $\hat{\tau}_b$ and hence the

513: eccentricity constraints (unless there are significant constraints

514: from RV observations or time of secondary transit; see \S\ref{SecRV}).

515: From the perspective of stellar modeling, the physical properties of a

516: star (e.g., radius, and density) are a function of at least three key

517: variables: stellar mass, composition, and age.  Hence stellar modeling

518: can only provide powerful constraints on the star's physical

519: properties if there are at least three observational constraints.  The

520: effective temperature ($T_{\rm eff}$) and metallicity ($[Fe/H]$) can

521: be accurately derived from a single high-precision spectroscopic

522: observation.  When a precise parallax ($\pi$) is available, the

523: stellar luminosity ($L_{\star}$) can be calculated from the apparent

524: magnitude (V; and a bolometric correction).  The three constraints

525: ($T_{\rm eff}$, $[Fe/H]$, $L_{\star}$) can be combined with stellar

526: evolution tracks to determine the stellar mass, radius, density,

527: etc. (e.g., Ford et al.\ 1999), with some well-known degeneracies

528: (e.g., hook region).  For relatively bright and nearby stars in the

529: California and Carnegie Planet Search and Hipparcos catalog, this

530: method has been used to estimate stellar parameters such as the mass

531: and radius (Valenti \& Fischer 2005; Takeda et al.\ 2007).  We applied

532: the Bayesian stellar parameter estimation code of Takeda et al.\ (2007) to

533: calculate density directly from the joint posterior distribution,

534: accounting for correlations between parameters and the non-linear

535: transformation between observable and derived quantities.  We find

536: that this method can determine the mean stellar density with an

537: accuracy of $\simeq 4-10\%$, with random uncertainties dominated by

538: the parallax.

539: 

540: We caution that there may also be systematic uncertainties due to the

541: stellar evolutionary tracks.  Such systematic effects could lead to

542: $\tau_o$ or $\tau_b$ being systematically over- or under-estimated and

543: that the error could be highly correlated with the stellar type or

544: other stellar parameters.  Ideally, such systematics would be

545: mitigated if the densities of stellar models were validated by

546: independent observations such as double-lined eclipsing binaries for

547: similar type stars.  Both the CoRoT and Kepler missions will

548: contribute to stellar astronomy and are likely to contribute towards

549: improving and testing the precision of stellar models.  In the absense

550: of externally validated models, systematic uncertainties in stellar

551: models may limit this techinque for systems with small eccentricities.

552: Fortunately, the stellar density enters only to the one third power,

553: so potential systematic effects are unlikely to be significant for

554: most planets that have sizable eccentricities.  In practice, the

555: eccentricity analysis should be coupled to a sensitivity analysis that

556: tests whether conclusions are sensitive to the choice of stellar models.

557: 

558: Unfortunately, many target stars for Kepler will be too faint and

559: distant to have well determined parallaxes.  One alternative approach

560: is to replace $L_{\star}$ with a spectroscopically determined stellar

561: surface gravity ($\log g$).  Unfortunately, $\log g$ is quite

562: difficult to measure precisely.  For a high-quality spectroscopic

563: observation (e.g., $\sigma_{T_{\rm eff}}\sim 100$K and $\sigma_{\rm

564: [Fe/H]}\sim 0.1$dex), the formal uncertainty in $\log g$ would

565: typically be $\sim 0.1$dex.  Unfortunately, there are typically

566: significant correlations with both $T_{\rm eff}$ and $[Fe/H]$ (Valenti

567: \& Fischer 2005).  For a main sequence solar-type star, the uncertainties in 

568: the atmophseric parameters would

569: translate to an uncertainty in $\rho_\star^{1/3}$ (and hence $\tau_b$)

570: of $\simeq 8$\%.  In some cases, additional constraints on stellar

571: properties such as rotation rate, strength of Ca HK emission, presence

572: of Li, and/or asteroseismology may be able to further constrain

573: stellar parameters and improve the stellar sensitivity for measuring

574: eccentricities.

575: 

576: 

577: \section{Statistical Methodology}

578: \label{SecStats}

579: %

580: In this section, we outline a few of the statistical approaches that

581: could be applied to translate measurements of $\tau_o$ and/or

582: $\tau_b$ into eccentricity constraints.  First, we outline a Bayesian

583: approach to calculating the joint posterior probability distribution

584: for the eccentricity and argument of periastron for individual

585: planets.  These constraints would be particularly valuable when combined with

586: additional constraints on dynamical properties of the system.

587: %

588: Second, we show that the distribution of normalized transit durations

589: contains significant information about the eccentricity distribution

590: of a population of planets.  

591: 

592: \subsection{Individual Planets}

593: \label{SecStatsMcmc}

594: %

595: For planets with high signal-to noise transit light curves and

596: well-measured stellar properties, we suggest a Bayesian framework for

597: determining the constraints on the eccentricity and argument of

598: pericenter.  To illustrate this method, we adopt non-informative

599: prior probability distributions of $p(e)=1$ for $0\le e<1$, $p(\omega)=1/(2\pi)$ for $0\le \omega < 1$, and

600: $p(\cos i)=1/2$ for $-1 \le \cos i < 1$.  We assume that the observed $\hat{\tau}_b$ is

601: normally distributed about the true value of $\tau_b$ (i.e.,

602: $\hat{\tau}_b \sim N(\tau_b,\sigma^2_{\tau_b}))$ with

603: $\sigma_{\tau_b}$ reflecting the uncertainty in the stellar parameters.

604: In Fig.\ \ref{FigTauContours}, we show examples of the joint posterior

605: probability distribution for six possible values of $\hat{\tau}_b$ .

606: Note that in this figure, $\omega$ is measured relative to the line of sight,

607: not relative to the plane of the sky, as is typical when using radial velocity observations.

608: For actual systems, Markov chain Monte Carlo based simulation methods

609: (Ford 2005, 2006) could be used to calculate posterior probability

610: distributions allowing for parameters with correlated uncertainties

611: and/or non-trivial error distributions (e.g., Holman et al.\ 2006;

612: Burke et al.\ 2007; Takeda et al.\ 2007), as well as eccentric orbits.

613: 

614: 

615: \subsection{Characterizing the Eccentricity Distribution of Transiting Planets}

616: \label{SecStatsKs}

617: %

618: We have performed Monte Carlo simulations to calculate the expected

619: distribution of transit durations for various eccentricity

620: distributions.  Here, we consider two limiting cases corresponding to

621: low and high signal-to-noise observations: a) only the transit depth,

622: duration, and orbital period are measured, so that the observed light

623: curve provides no information about the impact parameter (Fig.\

624: \ref{fig_hist_wob}), and b) the time of all four points of contact are

625: measured so that the observed light curve provides a good estimate of

626: the impact parameter (Fig.\ \ref{fig_hist_wb}).

627: 

628: 

629: \subsubsection{Impact Parameter Not Measured}

630: \label{SecStatsKsLow}

631: %

632: While it would be most desirable to measure the eccentricity for each

633: individual planet, this will not always be possible.  For some planets

634: (particularly those around faint stars), the light curve will not be

635: measured with sufficient precision to measure the orbital inclination

636: and sufficient high-precision RV observations will be impractical.  In

637: these cases, it will still be possible to constrain the mean

638: eccentricity of a population of such planets.  Since these challenges

639: will be most common for faint stars, we expect that missions such as

640: CoRoT and Kepler will detect a large number of such planets available

641: for statistical analyses.

642: 

643: We consider the distribution of $\tau_o\equiv t_{D} / t_{D,o}$, where

644: $t_D$ is the actual transit duration, $t_{D,o}$ is the transit

645: duration expected for the same planet and star, but assuming a

646: circular orbit ($e=0$) and central transit ($b=0$).  The distribution

647: for the impact parameter, $b$, is determined by assuming that the

648: inclination is distributed isotropically (subject to the constraint

649: that a transit occurs). We also assume that the observed

650: $\hat{\tau}_o$ is normally distributed about the true value of

651: $\tau_o$ (i.e., $\hat{\tau}_o \sim N(\tau_o,\sigma^2_{\tau_o}))$ with

652: $\sigma_{\tau_o}\simeq0.05-0.15$ due to uncertainty in the stellar

653: parameters (depending on the stellar properties and available

654: follow-up observations).

655: %

656: In Fig.\ \ref{fig_hist_wob}, we show the distribution of $\tau_o$ for

657: six eccentricity distributions.  

658: %

659: In Fig.\ \ref{fig_moments} (right), 

660: we plot the mean, standard deviation,

661: and skewness of $\tau_o$ for an ensemble of planets with a

662: given eccentricity.

663: %

664: For planets with a circular orbit and an isotropic distribution of

665: inclinations, $\tau_o$ will have a mean of $\simeq0.79$ and a standard

666: deviation of $\simeq0.22$ (Fig.\ \ref{fig_hist_wob}, panel a).

667: %

668: If we assume a uniform distribution of eccentricities ($e\sim U[0,1)$), then the mean $\tau_o$ decreases to 0.64 and the standard

669: deviation increases to 5.2.

670: %

671: Alternatively, if we instead assume an eccentricity distribution

672: similar to that observed for giant planets ($e\sim R(0.3)$, a Rayleigh

673: distribution with Rayleigh parameter 0.3; Juric \& Tremaine 2007),

674: then the mean decreases to $\simeq0.74$ and the standard deviation

675: increases to $\simeq0.29$.

676: %

677: Therefore, it would be possible to distinguish between all three

678: models with a plausible sample of planets, assuming normally

679: distributed errors and asymptotic scalings for the mean of the

680: distribution, even without measuring any impact parameters or RVs.

681: If we were to use only the mean value of $\tau_o$, then we

682: would be ignoring the variances and shape of the distribution.

683: Instead, a Komogorov-Smirnov test can be used to compare the observed

684: distribution of $\tau_o$ to the $\tau_o$ distribution predicted by a

685: given model.  

686: 

687: To determine how large an observed sample is needed to obtain

688: statistically significant results, we perform Monte Carlo simulations.

689: For each trial, we generate two samples of transiting planets.  The

690: first sample represents a hypothetical observed sample of $N_{pl}$

691: planets.  The second sample is much larger and is used to calculate

692: the predicted distribution of $\tau_o$ for a given theoretical model.

693: For both samples, we assume an isotropic distribution of viewing

694: angles, i.e., $\cos i \sim U[-1,1)$ and $\omega \sim U[0,2\pi)$. For

695: each trial, we test whether a Kolmogorov-Smirnov test can reject the

696: null hypothesis that the two distributions of $\tau_o$ are drawn from

697: a common distribution.  For each $N_{pl}$ we perform several thousand

698: trials; we increase $N_{pl}$ until we find that the null hypothesis is

699: rejected at the $95\%$ confidence level for at least half of the

700: trials.  We present our results in Table \ref{tab_one}.  Our

701: simulations show that a population of planets on circular orbits ($e =

702: 0$) could be distinguished from either $e\sim U[0,1)$ (eccentricities

703: distributed uniformly between zero and unity) or $e\sim R(0.3)$ (a

704: Rayleigh distribution with Rayleigh parameter of 0.3) by using

705: measurements of $\tau_o$ for $N_{pl}\go30$ planets, even without

706: measuring any impact parameters or RVs.  Distinguishing an

707: observed population with eccentricities distributed as $e\sim U[0,1)$

708: from a model with $e\sim R(0.3)$, would require $N_{pl}\go 117$ planets.

709: Distinguishing between an observed population with eccentricities

710: distributed as $e\sim R(0.3)$ rather than a model with $e\sim U[0,1)$, would

711: require $N_{pl}\go 174$ planets.  The difference is due to the fact

712: that we use a two-sample Kolmogorov-Smirnov test and the sample size

713: for the theoretical model is at least an order of magnitude greater

714: than the sample size for the sample of ``observed'' planets.

715: 

716: 

717: \subsubsection{Impact Parameter Measured}

718: \label{SecStatsKsHi}

719: %

720: When high quality photometry measures the impact parameter ($b$), then

721: we can analyze the distribution of $\tau_b\equiv t_{D} / t_{D,b}$,

722: where $t_{D,b}$ is the transit duration expected for the same planet

723: with the same impact parameter, but assuming a circular orbit.  

724: %

725: If all transiting planets had a single eccentricity, then the expected

726: mean value of $\tau_b$ is given by $\left<\tau_b\right>\simeq

727: \sqrt{1-e^2}$ (Tingley \& Sackett 2005) and the expected standard

728: deviation is $\left<\sigma_{\tau_b}\right> \simeq

729: (1-e^2)^{3/8}\sqrt{1-(1-e^2)^{1/4}}$.

730: %

731: We show the distribution of $\tau_b$ for several eccentricity

732: distributions in Fig.\ \ref{fig_hist_wb} and plot the mean, standard

733: deviation, and skewness for sample distributions of $\tau_b$ in Fig.\

734: \ref{fig_moments} (left).  Note that for small eccentricities, the

735: distribution of $\tau_b$ is strongly peaked, making it possible to

736: perform a significant test of the null hypothesis that terrestrial

737: planets have nearly circular orbits with a small sample size

738: (\S\ref{SecStatsNull}).  To account for the uncertainty in the stellar

739: parameters, we assume $\sigma_{\hat{\tau}_b}\simeq 0.05-0.15$ and

740: normally distributed errors and asymptotic scalings.  Then, it would

741: be possible to distinguish between various models for the eccentricity

742: distribution of observed planets with a few tens of transiting planets

743: (see Table \ref{tab_two}).  We find that the distribution of $\tau_b$

744: provides significant constraints on both the mean and the width of the

745: eccentricity distribution.  For example, Monte Carlo simulations using

746: the Kolmogorov-Smirnov test (similar to those described in

747: \S\ref{SecStatsKsLow}, but replacing $\tau_o$ with $\tau_b$) show that

748: distinguishing between two Gaussian eccentricity distributions with

749: mean eccentricities of 0.18 and 0.36 would require $N_{pl}\go30$

750: planets.  Similarly, distinguishing between two Gaussian eccentricity

751: distributions with a common mean of 0.38 and standard deviations of

752: 0.10 and 0.24 would require $N_{pl}\go 65$ planets based on a

753: Kolmogorov-Smirnov test and a 95\% confidence level.  However, it will

754: be very difficult to measure the higher order moments of the

755: eccentricity distribution (see Fig.\ \ref{FigDistribComp}).  In Table

756: \ref{tab_two}, we list the number of planet required for several

757: additional pairs of eccentricity distributions.

758: 

759: \subsection{Testing the Null Hypothesis of Circular Orbits for Individual Planets}

760: \label{SecStatsNull}

761: %

762: We are particularly interested in addressing the question, ``How

763: frequently are terrestrial planets on nearly circular orbits?''

764: Therefore, we suggest two statistical tests of the null-hypothesis

765: that each planet is on a circular orbit.  

766: 

767: \subsubsection{Low S/N Light Curves \& Long Duration Transits}

768: %

769: For an eccentric orbit with the transit near apocenter, the transit

770: duration can be significantly greater than would be expected for the

771: observed orbital period.  It is possible to measure a minimum

772: eccentricity for long transits, even for low S/N transits and without

773: measuring RV parameters.  Since a non-zero impact parameter can only

774: decrease the duration of transit (relative to a central transit), it

775: will be possible to reject the null hypothesis of a circular orbit

776: (and hence detect a non-zero eccentricity) for planets with long

777: duration transits such that $\tau_o$ or $\tau_b$ greater than unity.

778: Unfortunately, the geometric probability of a transit occurring is

779: greater for short transits when the planet is closer to the star.

780: %

781: Therefore, if eccentric planets are common, then there will be more

782: transits with $\tau<1$.  Fortunately, even a small number of long

783: transits can provide significant constraints for the eccentricity

784: distribution.  

785: 

786: 

787: \subsubsection{High S/N Light Curves \& Stellar Models}

788: 

789: For short-period planets that can be assumed to have

790: tidally circularized, $\tau_b =1$, so high precision light curve

791: observations may provide additional constraints on the star's

792: properties.  For example, Holman et al.\ (2007) measure

793: $\rho_{\star}^{1/3}$ to $\simeq 1.5$\% accuracy for TrES-2 (V=11.4).

794: When combined with theoretical models, this allows for very accurate

795: determinations of the stellar and planet radii (Sozzetti et al.\

796: 2007). 

797: This will enable tests of the null-hypothesis that a

798: planet is on a nearly circular orbit.  If it is accurate, then the

799: stellar properties would be measured quite precisely and all

800: measurements should be self-consistent.  On the other hand, if the

801:  planet is actually eccentric, then assuming a circular orbit could

802: result in an inconsistency.

803: %

804: For example, by comparing the allowed stellar models to a

805: spectroscopic measurement of $\log g$ one could recognize planets with

806: eccentricities exceeding $\simeq0.15$ for solar-like stars.

807: Alternatively, the putative location of the star in the ($T_{\rm

808: eff}$, $\rho_{\star}$, [Fe/H]) parameter space could be inconsistent

809: with any stellar model (e.g., Sozzetti et al.\ 2007).  The sensitivity

810: of this test will vary significantly with $T_{\rm eff}$ and be most

811: powerful for stars slightly cooler than the Sun.  Unfortunately, this

812: method is unlikely to be effective for planets with orbital periods

813: larger than a few days, since they may have eccentric orbits.

814: 

815: 

816: \subsection{Role of Radial Velocity Observations}

817: \label{SecRV}

818: %

819: The realization that many giant planets are on eccentric orbits was

820: the result of RV surveys (e.g., Butler et al.\ 2006 and references

821: therein).  Clearly, RV observations can measure orbital

822: eccentricities, provided that there is a sufficient number, phase coverage, and time

823: span of observations with sufficiently high precision (Ford 2005).  In

824: fact, the observational constraints from radial velocity observations

825: and the photometric method that we describe are complimentary.  Our

826: photometric method is most sensitive for measuring eccentricities of

827: planets if the pericenter direction is pointing towards or away from

828: the observer (Fig.\ \ref{FigTauContours}).  On the other hand, the

829: radial velocity method is most sensitive to measuring eccentricities

830: of planets with pericenter nearly in the plane of the sky (Ford 2005;

831: Laughlin et al.\ 2005).

832: 

833: Unfortunately, many of the CoRot and Kepler target stars (V$\sim9-16$)

834: will be much fainter than the typical targets of RV surveys

835: (V$\sim5-9$).  Ground-based transit surveys have obtained follow-up RV

836: observations for candidate planets, but typically at a relatively

837: modest precision.

838: %

839: While there are plans to obtain RV confirmation of planet candidates

840: identified by CoRoT or Kepler, these observations will require a

841: large investment of telescope time, especially when attempting to

842: detect Earth-mass or even Neptune-mass planets around relatively faint

843: stars.

844: %

845: Even when RV observations can confirm terrestrial planet candidates

846: found by CoRoT or Kepler, they will likely make use of the known orbital

847: period and phase to observe at near the extrema of the RV

848: curve.  Measuring the eccentricity requires measuring the shape of the

849: curve and hence many observations spread across a broad range of orbital

850: phases. 

851: %

852: The amplitude of the deviations of the RV curve from that expected for

853: the same planet on a circular orbit is less than the total RV

854: amplitude by a factor of $e$ (to lowest order in

855: eccentricity). Assuming a single planet with a orbital period known

856: from photometry, a detection of eccentricity for a planet with

857: $e\simeq 0.3$ will typically require an order of magnitude more

858: observing time than would be required to detect the planet's total RV

859: amplitude (at the same level of significance and assuming observations

860: nearly evenly distributed in orbital phase).  As an additional

861: complication, the RV method measures the reflex velocity of the star

862: induced by all planets, not just one transiting planet.  If a star

863: harbors a (potentially undetected) non-transiting planet with a RV

864: amplitude comparable to or greater than the radial velocity amplitude due

865: to the epicyclic motion of the transiting planet, then RV observations

866: would face an even greater challenge in measuring the transiting

867: planet's eccentricity.

868: %

869: Despite these challenges, we certainly encourage radial velocity

870: observations for constraining eccentricities whenever they are practical.

871: 

872: 

873: 

874: \section{Discussion}

875: \label{SecDiscuss}

876: %

877: We have described how photometric observations can constrain the

878: eccentricities of individual transiting planets and characterize the

879: eccentricity distribution of a population of planets.  For each

880: planet, we can test the null hypothesis that it is on a circular orbit

881: and calculate the posterior probability distribution for the

882: eccentricity.  A combination of such analyses for several transiting planets

883: could be used to characterize the eccentricity distribution of a

884: population of planets.  For example, this method could be used to

885: investigate how the fraction of eccentric orbits varies with the

886: orbital period or physical proprieties of the star and planet.  

887: We expect this type of analysis to become increasingly powerful given

888: the rapidly growing number of known transiting planets. 

889: 

890: This type of analysis will be particularly valuable for low-mass

891: transiting planets or transiting planets around faint host stars.

892: In both cases, radial velocity follow-up observations will be

893: extremely challenging.  This will be the case for many transiting

894: planet candidates found by space-based transit searches, such as CoRoT

895: and Kepler.  The capability of these missions to discover

896: terrestrial-mass planets is particularly exciting.  Our method could

897: could determine if terrestrial planets with low eccentricities like

898: the Earth are common or rare.  This would provide significant constraints

899: on theories proposed to explain the eccentricities of extrasolar planets.

900: For example, Kepler

901: might find many terrestrial planets with low eccentricity orbits,

902: suggesting that the mechanisms that excite the eccentricities of giant

903: planets are often ineffective for terrestrial mass planets.  In this

904: scenario, those terrestrial planets that do have large eccentricities might

905: typically be accompanied by nearby giant planets, suggesting that it

906: is the giant planets are responsible for exciting the eccentricities

907: of terrestrial planets (Veras \& Armitage 2005, 2006).  Alternatively,

908: Kepler might find that terrestrial planets are much more common than

909: giant planets, and yet they still commonly have large eccentricities.

910: This could arise from eccentricity excitation mechanisms that do not

911: require giant planets, or due to interactions with previous giant

912: planets that have since been ejected, accreted or destroyed by the star (e.g.,

913: Ford et al.\ 2005; Raymond et al.\ 2006; Mandell et al.\ 2007).

914: 

915: Our method also provides a means for studying the tidal evolution of

916: short-period planets.  Tidal effects are likely to circularize planets with sufficiently

917: short orbital periods.  For short-period planets, we can compute the tidal

918: circularization timescale ($t_{\rm circ}$) based on the properties of

919: the star and planet.  If there is a sharp transition between circular

920: and eccentric orbits, then this could be used to place constraints on

921: tidal theory (e.g., Zahn \& Bouchet 1989; Melo et al.\ 2001; Mathieu

922: et al.\ 2004).  For eccentric planets with relatively short $t_{\rm

923: circ}$, it may be possible to place a lower limit on the $Q$ factor

924: the is related to the planet's internal structure (e.g., Ford et al.\

925: 1999; Bodenheimber et al.\ 2001; Maness et al.\ 2007; Mardling 2007).

926: 

927: In a Bayesian framework, one can calculate the

928: posterior probability distribution for the fraction of each orbit

929: during which the planet-star separation is between an inner and outer

930: cut-off.  If the cut-offs are set to be the putative boundary of the

931: habitable zone, then we can then ask, ``What fraction of terrestrial

932: planets are in the habitable zone for some/at least half/all of their

933: orbit?''.  The results of such investigations could have implications

934: for the climates of potentially habitable planets, the frequency of

935: such planets, and the design of future missions that aim to detect and

936: characterize nearby Earth-like planets that could harbor life (e.g.,

937: Marcy et al.\ 2005).

938: 

939: We have demonstrated that it is practical to collect sufficient

940: photons to characterize the eccentricity distribution of terrestrial

941: extrasolar planets, but we assumed that limb-darkening parameters can

942: be well constrained by some combination of stellar modeling and

943: external observations.  Our analytical estimates have not incorporated

944: limb darkening effects or potential systematic uncertainties in

945: stellar models.  Future search should address both of these effects.

946: Multi-wavelength observations (particularly in the infrared) could be

947: particularly useful for addressing both these issues.  In particular,

948: we plan to investigate the potential for combinations of space-based

949: detections and ground-based follow-up observations to improve the

950: characterization of the eccentricities of transiting planets.

951: 

952: Finally, we note that this method for characterizing the

953: eccentricities of terrestrial planets from transit light curves

954: underscores the importance of developing and validating precise and

955: accurate stellar models.  Uncertainties in stellar parameters models

956: are expected to dominate the error budget for bright target stars.

957: The potential for systematic uncertainties due to stellar modeling

958: will make it particularly challenging to study low eccentricity

959: systems as a function of stellar properties.  Fortunately, we these

960: concerns would not preclude our method from being applied to

961: terrestrial-sized planets recognizing the relatively large

962: eccentricities typical for giant planets.

963: 

964: \acknowledgements We thank G. Bakos, D.\ Charboneau, M.\ Holman, D.\

965: Latham, G.\ Takeda, and an anonymous referee for discussions and

966: suggestions.

967: %

968: Support for E.B.F.\ was provided by NASA through Hubble Fellowship

969: grant HST-HF-01195.01A awarded by the Space Telescope Science

970: Institute, which is operated by the Association of Universities for

971: Research in Astronomy, Inc., for NASA, under contract NAS 5-26555.

972: Additional support for E.B.F.\ and D.V.\ was provided by the

973: University of Florida.

974: 

975: \clearpage

976: \begin{thebibliography}{}

977: 

978: 

979: %\bibitem[Armitage et al.(2002)]{2002MNRAS.334..248A} Armitage, P.~J., Livio, M., Lubow, S.~H., \& Pringle, J.~E.\ 2002, \mnras, 334, 248

980: 

981: 

982: \bibitem[Artymowicz(1992)]{1992PASP..104..769A} Artymowicz, P.\ 1992, 

983: \pasp, 104, 769 

984: 

985: \bibitem[Bakos et al.(2007)]{2007arXiv0705.0126B} Bakos, G.~A., et al.\ 

986: 2007, ArXiv e-prints, 705, arXiv:0705.0126 

987: 

988: \bibitem[Barnes(2007)]]{2008arXiv0708.0243} Barnes, J.W. 2007, ArXiv e-prints, 807, arXiv:0708.0243

989: 

990: \bibitem[Basri et al.(2005)]{2005NewAR..49..478B} Basri, G., Borucki, 

991: W.~J., \& Koch, D.\ 2005, New Astronomy Review, 49, 478 

992: 

993: %\bibitem[Bedding et al.(2005)]{2005A&A...432L..43B} Bedding, T.~R., et al.\  2005, \aap, 432, L43 

994: 

995: 

996: \bibitem[Beer et al.(2004)]{2004MNRAS.354..763B} Beer, M.~E., King, A.~R., 

997: Livio, M., \& Pringle, J.~E.\ 2004, \mnras, 354, 763 

998: 

999: 

1000: \bibitem[Bodenheimer et al.(2001)]{2001ApJ...548..466B} Bodenheimer, P., 

1001: Lin, D.~N.~C., \& Mardling, R.~A.\ 2001, \apj, 548, 466 

1002: 

1003: 

1004: \bibitem[Brown et al.(2001)]{2001ApJ...552..699B} Brown, T.~M., 

1005: Charbonneau, D., Gilliland, R.~L., Noyes, R.~W., \& Burrows, A.\ 2001, 

1006: \apj, 552, 699 

1007: 

1008: 

1009: \bibitem[Burke(2008)]{2008arXiv0801.2579} Burke, C.~J. 2008, ArXiv e-prints, 801, arXiv:0801.2579

1010: 

1011: \bibitem[Burke et al.(2007)]{2007arXiv0705.0003B} Burke, C.~J., et al.\ 

1012: 2007, ArXiv e-prints, 705, arXiv:0705.0003   

1013: 

1014: 

1015: \bibitem[Butler et al.(2006)]{2006ApJ...646..505B} Butler, R.~P., et al.\ 

1016: 2006, \apj, 646, 505 

1017: 

1018: 

1019: \bibitem[Chatterjee et al.(2007)]{2007astro.ph..3166C} Chatterjee, S., 

1020: Ford, E.~B., \& Rasio, F.~A.\ 2007, ArXiv Astrophysics e-prints, 

1021: arXiv:astro-ph/0703166 

1022: 

1023: 

1024: \bibitem[Claret(2000)]{2000A&A...363.1081C} Claret, A.\ 2000, \aap, 363, 

1025: 1081 

1026: 

1027: 

1028: \bibitem[Fischer \& Valenti(2005)]{2005ApJ...622.1102F} Fischer, D.~A., \& 

1029: Valenti, J.\ 2005, \apj, 622, 1102 

1030: 

1031: 

1032: % \bibitem[Ford \& Rasio(2005)]{2005prpl.conf.8360F} Ford, E.~B., \& Rasio, F.~A.\ 2005, Protostars and Planets V, 8360 

1033: 

1034: 

1035: %\bibitem[Ford et al.(2003)]{2003ASPC..294..181F} Ford, E.~B., Rasio, F.~A., \& Yu, K.\ 2003, Scientific Frontiers in Research on Extrasolar Planets, 294, 181 

1036: 

1037: 

1038: \bibitem[Ford et al.(2000)]{2000ApJ...535..385F} Ford, E.~B., Kozinsky, B., 

1039: \& Rasio, F.~A.\ 2000, \apj, 535, 385 

1040: 

1041: \bibitem[Ford(2006)]{2006ApJ...642..505F} Ford, E.~B.\ 2006, \apj, 642, 505 

1042: 

1043: 

1044: \bibitem[Ford(2005)]{2005AJ....129.1706F} Ford, E.~B.\ 2005, \aj, 129, 1706 

1045: 

1046: 

1047: %\bibitem[Ford(2004)]{2004astro.ph.12703F} Ford, E.~B.\ 2004, ArXiv Astrophysics e-prints, arXiv:astro-ph/0412703 

1048: 

1049: 

1050: \bibitem[Ford \& Gaudi(2006)]{2006ApJ...652L.137F} Ford, E.~B., \& Gaudi, 

1051: B.~S.\ 2006, \apjl, 652, L137 

1052: 

1053: 

1054: %\bibitem[Ford \& Gregory(2006)]{2006astro.ph..8328F} Ford, E.~B., \& Gregory, P.~C.\ 2006, ArXiv Astrophysics e-prints, arXiv:astro-ph/0608328 

1055: 

1056: 

1057: %\bibitem[Ford et al.(2001)]{2001Icar..150..303F} Ford, E.~B., Havlickova, M., \& Rasio, F.~A.\ 2001, Icarus, 150, 303 

1058: 

1059: 

1060: \bibitem[Ford et al.(2005)]{2005Natur.434..873F} Ford, E.~B., Lystad, V., \& Rasio, F.~A.\ 2005, \nat, 434, 873 

1061: 

1062: 

1063: \bibitem[Ford \& Rasio(2007)]{2007astro.ph..3163F} Ford, E.~B., \& Rasio, 

1064: F.~A.\ 2007, ArXiv Astrophysics e-prints, arXiv:astro-ph/0703163 

1065: 

1066: 

1067: %\bibitem[Ford \& Rasio(2006)]{2006ApJ...638L..45F} Ford, E.~B., \& Rasio, F.~A.\ 2006, \apjl, 638, L45 

1068: 

1069: 

1070: \bibitem[Ford et al.(1999)]{1999ApJ...514..411F} Ford, E.~B., Rasio, F.~A., 

1071: \& Sills, A.\ 1999, \apj, 514, 411 

1072: 

1073: 

1074: \bibitem[Gaidos \& Williams(2004)]{2004NewA...10...67G} Gaidos, E., \& 

1075: Williams, D.~M.\ 2004, New Astronomy, 10, 67 

1076: 

1077: 

1078: \bibitem[Goldreich \& Sari(2003)]{2003ApJ...585.1024G} Goldreich, P., \& 

1079: Sari, R.\ 2003, \apj, 585, 1024 

1080: 

1081: 

1082: \bibitem[Holman et al.(1997)]{1997Natur.386..254H} Holman, M., Touma, J., 

1083: \& Tremaine, S.\ 1997, \nat, 386, 254 

1084: 

1085: \bibitem[Holman et al.(2006)]{2006ApJ...652.1715H} Holman, M.~J., et al.\ 

1086: 2006, \apj, 652, 1715 

1087: 

1088: 

1089: \bibitem[Holman et al.(2007)]{2007arXiv0704.2907H} Holman, M.~J., et al.\ 

1090: 2007, ArXiv e-prints, 704, arXiv:0704.2907 

1091: 

1092: 

1093: \bibitem[Jha et al.(2000)]{2000ApJ...540L..45J} Jha, S., Charbonneau, D., 

1094: Garnavich, P.~M., Sullivan, D.~J., Sullivan, T., Brown, T.~M., \& Tonry, 

1095: J.~L.\ 2000, \apjl, 540, L45 

1096: 

1097: \bibitem[Juric \& Tremaine(2007)]{2007astro.ph..3160J} Juric, M., \& 

1098: Tremaine, S.\ 2007, ArXiv Astrophysics e-prints, arXiv:astro-ph/0703160 

1099: 

1100: 

1101: \bibitem[Kasting et al.(1993)]{1993Icar..101..108K} Kasting, J.~F., 

1102: Whitmire, D.~P., \& Reynolds, R.~T.\ 1993, Icarus, 101, 108 

1103: 

1104: 

1105: \bibitem[Laughlin \& Adams(1998)]{1998ApJ...508L.171L} Laughlin, G., \& 

1106: Adams, F.~C.\ 1998, \apjl, 508, L171 

1107: 

1108: 

1109: \bibitem[Laughlin et al.(2005)]{2005ApJ...629L.121L} Laughlin, G., Marcy, 

1110: G.~W., Vogt, S.~S., Fischer, D.~A., \& Butler, R.~P.\ 2005, \apjl, 629, 

1111: L121 

1112: 

1113: 

1114: %\bibitem[Livio \& Pringle(2003)]{2003MNRAS.346L..42L} Livio, M., \& Pringle, J.~E.\ 2003, \mnras, 346, L42 

1115: 

1116: 

1117: \bibitem[Mandel \& Agol(2002)]{2002ApJ...580L.171M} Mandel, K., \& Agol, 

1118: E.\ 2002, \apjl, 580, L171 

1119: 

1120: 

1121: \bibitem[Mandell et al.(2007)]{2007ApJ...660..823M} Mandell, A.~M., 

1122: Raymond, S.~N., \& Sigurdsson, S.\ 2007, \apj, 660, 823 

1123: 

1124: 

1125: \bibitem[Maness et al.(2007)]{2007PASP..119...90M} Maness, H.~L., Marcy, 

1126: G.~W., Ford, E.~B., Hauschildt, P.~H., Shreve, A.~T., Basri, G.~B., Butler, 

1127: R.~P., \& Vogt, S.~S.\ 2007, \pasp, 119, 90 

1128: 

1129: 

1130: \bibitem[Marcy et al.(2005)]{2005ASPC..338..191M} Marcy, G.~W., Fischer, 

1131: D.~A., McCarthy, C., \& Ford, E.~B.\ 2005, Astrometry in the Age of the 

1132: Next Generation of Large Telescopes, 338, 191 

1133: 

1134: 

1135: \bibitem[Mardling(2007)]{2007MNRAS..382..1768} Mardling, R.~A. 2007, \mnras, 382, 1768

1136: 

1137: \bibitem[Mathieu et al.(2004)]{2004ApJ...602L.121M} Mathieu, R.~D., Meibom, 

1138: S., \& Dolan, C.~J.\ 2004, \apjl, 602, L121 

1139: 

1140: 

1141: \bibitem[Melo et al.(2001)]{2001A&A...378..898M} Melo, C.~H.~F., Covino, 

1142: E., Alcal{\'a}, J.~M., \& Torres, G.\ 2001, \aap, 378, 898 

1143: 

1144: 

1145: \bibitem[Namouni(2007)]{2007astro.ph..2203N} Namouni, F.\ 2007, ArXiv  Astrophysics e-prints, arXiv:astro-ph/0702203 

1146: 

1147: 

1148: \bibitem[Raymond et al.(2006)]{2006Sci...313.1413R} Raymond, S.~N., 

1149: Mandell, A.~M., \& Sigurdsson, S.\ 2006, Science, 313, 1413 

1150: 

1151: 

1152: %\bibitem[Ribas \& Miralda-Escud{\'e}(2007)]{2007A&A...464..779R} Ribas, I., \& Miralda-Escud{\'e}, J.\ 2007, \aap, 464, 779 

1153: 

1154: 

1155: %\bibitem[Robinson et al.(2006)]{2006ApJ...643..484R} Robinson, S.~E., Laughlin, G., Bodenheimer, P., \& Fischer, D.\ 2006, \apj, 643, 484 

1156: 

1157: 

1158: %\bibitem[Sari \& Goldreich(2004)]{2004ApJ...606L..77S} Sari, R., \& Goldreich, P.\ 2004, \apjl, 606, L77 

1159: 

1160: 

1161: \bibitem[Seager \& Mall{\'e}n-Ornelas(2003)]{2003ApJ...585.1038S} Seager, 

1162: S., \& Mall{\'e}n-Ornelas, G.\ 2003, \apj, 585, 1038 

1163: 

1164: \bibitem[Sozzetti et al.(2007)]{2007arXiv0704.2938S} Sozzetti, A., Torres, 

1165: G., Charbonneau, D., Latham, D.~W., Holman, M.~J., Winn, J.~N., Laird, 

1166: J.~B., \& O'Donovan, F.~T.\ 2007, ArXiv e-prints, 704, arXiv:0704.2938 

1167: 

1168: 

1169: \bibitem[Takeda et al.(2007)]{2007ApJS..168..297T} Takeda, G., Ford, E.~B., 

1170: Sills, A., Rasio, F.~A., Fischer, D.~A., \& Valenti, J.~A.\ 2007, \apjs, 

1171: 168, 297 

1172: 

1173: 

1174: \bibitem[Tingley \& Sackett(2005)]{2005ApJ...627.1011T} Tingley, B., \& 

1175: Sackett, P.~D.\ 2005, \apj, 627, 1011 

1176: 

1177: 

1178: \bibitem[Valenti \& Fischer(2005)]{2005ApJS..159..141V} Valenti, J.~A., \& 

1179: Fischer, D.~A.\ 2005, \apjs, 159, 141 

1180: 

1181: 

1182: \bibitem[Veras \& Armitage(2006)]{2006ApJ...645.1509V} Veras, D., \& 

1183: Armitage, P.~J.\ 2006, \apj, 645, 1509 

1184: 

1185: 

1186: \bibitem[Veras \& Armitage(2005)]{2005ApJ...620L.111V} Veras, D., \& 

1187: Armitage, P.~J.\ 2005, \apjl, 620, L111 

1188: 

1189: 

1190: \bibitem[Williams \& Pollard(2002)]{2002IJAsB...1...61W} Williams, D.~M., 

1191: \& Pollard, D.\ 2002, International Journal of Astrobiology, 1, 61 

1192: 

1193: 

1194: %\bibitem[Winn et al.(2007)]{2007AJ....133...11W} Winn, J.~N., Holman, M.~J., \& Fuentes, C.~I.\ 2007, \aj, 133, 11 

1195: 

1196: 

1197: %\bibitem[Winn et al.(2007)]{2007AJ....133.1828W} Winn, J.~N., et al.\ 2007, \aj, 133, 1828 

1198: 

1199: 

1200: \bibitem[Winn et al.(2007)]{2007ApJ...657.1098W} Winn, J.~N., Holman, 

1201: M.~J., \& Roussanova, A.\ 2007, \apj, 657, 1098 

1202: 

1203: 

1204: %\bibitem[Winn et al.(2005)]{2005ApJ...631.1215W} Winn, J.~N., et al.\ 2005, \apj, 631, 1215 

1205: 

1206: 

1207: \bibitem[Zahn \& Bouchet(1989)]{1989A&A...223..112Z} Zahn, J.-P., \& 

1208: Bouchet, L.\ 1989, \aap, 223, 112 

1209: 

1210: 

1211: \bibitem[Zakamska \& Tremaine(2004)]{2004AJ....128..869Z} Zakamska, N.~L., 

1212: \& Tremaine, S.\ 2004, \aj, 128, 869 

1213: 

1214: %

1215: \end{thebibliography}

1216: \clearpage

1217:               

1218: \begin{deluxetable}{ccccccccc}

1219: \rotate

1220: \tabletypesize{\small}

1221: \tablewidth{0pt}

1222: \tablecaption{Number of Planets Required to Distinguish between Distributions (Low S/N)\label{tab_one}}

1223: \tablehead{

1224: \colhead{ } & \colhead{$e=0$} & \colhead{$e\sim R(.1)$} & \colhead{$e\sim R(.2)$} & \colhead{$e\sim R(.3)$} & \colhead{$e\sim R(.4)$} & \colhead{$e\sim R(.5)$} & \colhead{$e\sim R(.6)$} & \colhead{$e\sim U[0,1)$}}

1225: \startdata

1226: $e=0$          & \nodata & 98 & 44 & 30 & 22 & 19 & 16 & 26 \\

1227: $e\sim R(.1)$  & 78 & \nodata & 187 & 53 & 33 & 23 & 20 & 34 \\

1228: $e\sim R(.2)$  & 51 & 171 & \nodata & 250 & 84 & 42 & 33 & 52 \\

1229: $e\sim R(.3)$  & 36 & 54 & 180 & \nodata & 355 & 117 & 67 & 174 \\

1230: $e\sim R(.4)$  & 25 & 29 & 68 & 270 & \nodata & 575 & 215 & 675 \\

1231: $e\sim R(.5)$  & 20 & 24 & 42 & 108 & 540 & \nodata & $>$1000 & 975 \\

1232: $e\sim R(.6)$  & 17 & 19 & 28 & 55 & 155 & 850 & \nodata & 255 \\

1233: $e\sim U[0,1)$ & 26 & 29 & 53 & 117 & 540 & 850 & 310 & \nodata \\

1234: \enddata

1235: %

1236: \tablenotetext{a}{We list the number of planets required to 

1237: distinguish an observed distribution (taken from the top row) from a

1238: theoretical distribution (taken from the left column) at the 95\%

1239: confidence level based on a Kolmogorov-Smirnov test applied to

1240: $\tau_o$.}

1241: %

1242: \end{deluxetable}

1243: 

1244: \begin{deluxetable}{ccccccccc}

1245: \rotate

1246: \tabletypesize{\footnotesize}

1247: \tablewidth{0pt}

1248: \tablecaption{Number of Planets Required to Distinguish between Distributions (High S/N)\label{tab_two}}

1249: \tablehead{

1250: \colhead{ } & \colhead{$e=0$} & \colhead{$e\sim R(.1)$} & \colhead{$e\sim R(.2)$} & \colhead{$e\sim R(.3)$} & \colhead{$e\sim R(.4)$} & \colhead{$e\sim R(.5)$} & \colhead{$e\sim R(.6)$} & \colhead{$e\sim U[0,1)$}}

1251: \startdata

1252: \cutinhead{$\sigma_{\tau_b} = 0.05$}

1253: $e=0$          & \nodata & 14 & 11 & 8 & 7 & 7 & 6 & 8 \\

1254: $e\sim R(.1)$  & 17 & \nodata & 47 & 20 & 12 & 9 & 9 & 16 \\

1255: $e\sim R(.2)$  & 11 & 45 & \nodata & 77 & 33 & 18 & 17 & 30 \\

1256: $e\sim R(.3)$  & 8 & 15 & 72 & \nodata & 158 & 58 & 34 & 108 \\

1257: $e\sim R(.4)$  & 7 & 9 & 21 & 122 & \nodata & 380 & 127 & 385 \\

1258: $e\sim R(.5)$  & 6 & 7 & 14 & 46 & 335 & \nodata & 725 & 275 \\

1259: $e\sim R(.6)$  & 5 & 6 & 11 & 28 & 99 & 495 & \nodata & 100 \\

1260: $e\sim U[0,1)$ & 8 & 11 & 20 & 67 & 355 & 305 & 133 & \nodata \\

1261: \cutinhead{$\sigma_{\tau_b} = 0.15$}

1262: $e=0$          & \nodata & 29 & 19 & 13 & 10 & 9 & 8 & 11 \\

1263: $e\sim R(.1)$  & 35 & \nodata & 73 & 23 & 17 & 11 & 12 & 22 \\

1264: $e\sim R(.2)$  & 19 & 65 & \nodata & 88 & 43 & 23 & 22 & 38 \\

1265: $e\sim R(.3)$  & 13 & 23 & 97 & \nodata & 210 & 70 & 47 & 125 \\

1266: $e\sim R(.4)$  & 8 & 14 & 31 & 145 & \nodata & 465 & 148 & 480 \\

1267: $e\sim R(.5)$  & 8 & 11 & 18 & 60 & 355 & \nodata & 775 & 375 \\

1268: $e\sim R(.6)$  & 7 & 8 & 13 & 37 & 120 & 545 & \nodata & 135 \\

1269: $e\sim U[0,1)$ & 10 & 14 & 26 & 75 & 410 & 475 & 175 & \nodata \\

1270: \enddata

1271: %

1272: \tablenotetext{a}{We list the number of planets required to

1273: distinguish an observed distribution (taken from the top row) from a

1274: theoretical distribution (taken from the left column) at the 95\%

1275: confidence level based on a Kolmogorov-Smirnov test applied to

1276: $\hat{\tau}_b$ with $\sigma_{\tau_b} = 0.15$ (bottom) and 0.05 (top).}

1277: %

1278: \end{deluxetable}

1279: 

1280: \clearpage

1281: %

1282: \begin{figure}

1283: \plotone{f1.eps} % TransitFig.eps

1284: \caption{Geometry of Transit: Here we illustrate the

1285: path (dashed line) of a planet (small circles) as it transits a star

1286: (large circle).  The vertical dotted lines connect the planet to the

1287: schematic light curve (solid curve) at the time of each of the four

1288: points of contact.  Measuring the times of all four points of contact

1289: allows for a measurement of the impact parameter ($b$) and

1290: significantly increases the precision of the constraint on the

1291: planet's eccentricity.  If only the total transit duration ($t_D$) is

1292: measured, then the eccentricity distribution can still be constrained

1293: for a larger population of transiting planets.

1294: %

1295: \label{FigGeo}}

1296: \end{figure}

1297: %

1298: 

1299: %

1300: \begin{figure}

1301: \plotone{f2.eps} % tauvsomega.eps}

1302: \caption{Transit Duration as a function of eccentricity ($e$) and argument of periastron ($\omega$):

1303: %

1304: The vertical axis shows $\tau_b$, the ratio of the actual transit

1305: duration to the transit duration for a similar planet on a circular

1306: orbit (with the same sizes, orbital period and impact parameter). The

1307: curves show the analytic approximation for the transit duration (Eqn.\

1308: 1) for four values of eccentricity, 0.25, 0.5, 0.75, 0.9.  The points

1309: are the exact durations for a planet on a Keplerian orbit.  For a

1310: planet on a circular orbit, $\tau_b=1$, but for eccentric planets

1311: $\tau_b$ can be larger (for planets that transit near apocenter) or

1312: smaller (for planets that transit near pericenter).  Thus, a

1313: measurement of the transit duration can be used to constrain a

1314: combination of $e$ and $\omega$ (measured from the direction of the

1315: observer), as well as to place a lower limit on the

1316: eccentricity.

1317: %

1318: \label{fig_tauvsomega}   }

1319: \end{figure}

1320: %

1321: 

1322: \begin{figure}

1323: \plotone{f3.eps} % TauProbFixedE.eps}

1324: \caption{

1325: Probability Distribution for Transit Duration at a given eccentricity ($e$):

1326: %

1327: Here we show the probability distribution (solid histogram) for

1328: $\tau_b$ for an ensemble of planets with a single fixed eccentricity,

1329: assuming a uniform distribution of the argument of pericenter and an

1330: isotropic distribution of inclinations.  The dotted curves are the

1331: cumulative probability distributions.  For large eccentricities, there

1332: is a small fraction of very long transits ($\tau_b>2$ off the scale).

1333: %

1334: \label{FigProbTau} }

1335: \end{figure}

1336: %

1337: 

1338: %

1339: \begin{figure}

1340: \plotone{f4.eps} % LimbDarkening.eps}

1341: \caption{Transit light curves illustrating effects of

1342: limb-darkening.

1343: %

1344: Here we show the observed intensity ($I(z)$; normalized to total

1345: stellar flux, $I_{\star}$) as a function of $z$, the projected

1346: separation from the center of the star (measured in stellar radii).

1347: The vertical dotted lines indicate the points of contact.  For

1348: this illustration, we adopt a

1349: quadratic limb-darkening model (Mandel \& Agol 2002) and limb

1350: darkening coefficients of $\gamma_1=0.4382$ and $\gamma_2=0.2924$

1351: based on ATLAS models for solar-like star in V band (Claret 2000). 

1352: %

1353: {\em Top:} For a Jupiter-sized planet, 

1354: %

1355: % the change in slope at the second/third point of contact is modest due to a combination of limb darkening and a relatively large planet to star size ratio. 

1356: %

1357: {\em Bottom:} For an Earth-sized planet.

1358: %

1359: %, the duration of ingress/egress is shorter, reducing the number of photons received during ingress/egress.  Fortunately, a smaller planet size results in a better measurement of the limb darkening parameters and thus improve the measurement of the times of the second/third point of contact.

1360: %

1361: \label{FigLimbDarkening}   }

1362: \end{figure}

1363: %

1364: 

1365: \begin{figure}

1366: \plotone{f5.eps} % keplercor.eps

1367: \caption{Correlations between transit duration and impact parameter or

1368: limb darkening coefficients.  Here we shows contour of constant

1369: $\Delta\chi^2$ = 1, 4, and 9 with $\hat{t}_D/t_D$ (the ratio of the

1370: model transit duration to the actual transit duration) on the

1371: $y$-axis.  The various collumns have $x$-axes of the impact parameter

1372: ($b$; left), the sum of the quadratic limb darkening coefficients

1373: ($u_1+u_2$; center), and the difference of the quadratic limb

1374: darkening coefficients ($u_1-u_2$; right).  The top row shows a 1

1375: $R_\oplus$ planet and the bottom row shows a $2 R_\oplus$ planet, both

1376: assumed to be at 1 AU from a solar mass star.

1377: %

1378: \label{FigContours} }

1379: \end{figure}

1380: 

1381: 

1382: 

1383: 

1384: %

1385: %

1386: \begin{figure}

1387: \plotone{f6.eps} % TauContours.eps}

1388: \caption{

1389: Posterior Joint Probability Distribution for $e$ and $\omega$:

1390:  Here we consider the eccentricity constraint based

1391: on the measured value of $\tau_b$ for an individual planet with a

1392: measured impact parameter.  We show contours equivalent to 1, 2,

1393: 3,...-$\sigma$ bounds on the combination of eccentricity and argument

1394: of pericenter.  The shaded regions are excluded by the measurement of

1395: $\tau_b$.  The panels correspond to $\tau_b=0.25$ (top left), 0.5

1396: (top center), 0.75 (top right), 1 (bottom left), 1.25 (bottom center),

1397: and 1.5 (bottom right).  Here we assume that the detection probability

1398: (given that a transit occurs) is independent of transit duration and

1399: that the measurement of $\hat{\tau}_b$ is normally distributed and has

1400: a standard deviation, $\sigma_{\tau_b}=0.1$.

1401: %

1402: \label{FigTauContours} }

1403: \end{figure}

1404: %

1405: 

1406: %

1407: %

1408: %\begin{figure}

1409: %\plottwo{hist_wob.eps}{hist_wb.eps}

1410: %\caption{Distribution of Transit Duration for Various Eccentricity Distributions:

1411: %

1412: %{\em Left:} Histograms of $\tau_o$, the ratio of the observed transit duration to that expected for the same planet, star, and orbital period, but a circular orbit and a central transit.  Panel a corresponds to only circular orbits, panels b-f correspond to a Gaussian eccentricity distribution with means of 0.25 (b), 0.5 (c), and 0.75 (d) and a standard deviation of 0.3.

1413: %

1414: %{\em Right:} Histograms of $\tau_b$, the ratio of the observed transit duration to that expected for a similar planet, star, impact parameter, and orbital period, but a circular orbit.  Panels b-f are for the same eccentricity distributions as on the left.  Panel a has been replaced with a Rayleigh distribution of eccentricities (with Rayleigh parameter 0.3).  For planets with only circular orbits, the histogram would be a delta function at $\tau_b=1$.

1415: %

1416: %\label{fig_hist_wb}

1417: %\label{fig_hist_wob}  

1418: %}

1419: %\end{figure}

1420: %

1421: %

1422: 

1423: %

1424: %

1425: \begin{figure}

1426: \plotone{f7.eps} % fig6a.eps}

1427: \caption{Distribution of Transit Duration for Various Eccentricity Distributions:

1428: %

1429: We show histograms (solid curves) and cumlative distributions (dashed curves) of $\tau_o$, the ratio of the observed transit duration to

1430: that expected for the same planet, star, and orbital period, but a

1431: circular orbit and a central transit.  Panel a corresponds to only

1432: circular orbits, panels b-f correspond to a Rayleigh eccentricity

1433: distribution with Rayleigh parameter of 0.1 (b), 0.2 (c), and 0.3 (d),

1434: 0.4 (e), and 0.5 (f).

1435: %

1436: %

1437: \label{fig_hist_wob}  

1438: }

1439: \end{figure}

1440: %

1441: %

1442: 

1443: %

1444: %

1445: \begin{figure}

1446: \plotone{f8.eps} % fig6b.eps}

1447: \caption{Distribution of Transit Duration for Various Eccentricity Distributions:

1448: %

1449: We show histograms (solid curves) and cumlative distributions (dashed curves) of $\tau_b$, the ratio of the observed transit duration to

1450: that expected for a similar planet, star, impact parameter, and

1451: orbital period, but a circular orbit.  Panel a corresponds to a

1452: uniform eccentricity distribution, panels b-f correspond to a Rayleigh

1453: eccentricity distribution with Rayleigh parameter of 0.1 (b), 0.2 (c),

1454: and 0.3 (d), 0.4 (e), and 0.5 (f).

1455: %

1456: %

1457: \label{fig_hist_wb}

1458: }

1459: \end{figure}

1460: %

1461: %

1462: 

1463: 

1464: %

1465: %

1466: \begin{figure}

1467: \plotone{f9.eps} % moments.eps}

1468: \caption{Moments of the Normalized Transit Duration Distribution:

1469:  Here we consider several ensembles of transiting

1470: planets, each with a single fixed eccentricity.  For each

1471: eccentricity, we calculate the distribution of $\tau_b$ (left; for

1472: transits with measured impact parameters) and $\tau_o$ (right; for

1473: transits without measured impact parameters).  Here we plot the mean,

1474: standard deviation, and skewness for both both $\tau_b$ and $\tau_o$

1475: for each eccentricity.  We show analytic approximations with curves

1476: when available.

1477: %

1478: \label{fig_moments} }

1479: \end{figure}

1480: %

1481: %

1482:  

1483: %

1484: %

1485: \begin{figure}

1486: \plotone{f10.eps} % hist.n.r3.eps}

1487: \caption{Distribution of Transit Duration for Eccentricity Distributions with Common Mean and Variance:

1488: %

1489: {\em Top:} Histograms of $\tau_o$, the ratio of the observed transit

1490: duration to that expected for the same planet, star, and orbital

1491: period, but a circular orbit and a central transit.  We show results

1492: for two eccentricity distributions: normal (solid) and Rayleigh

1493: (dashed).  The mean and variance of the normal distribution have been

1494: chosen to match that of a Rayleigh distribution with Rayleigh

1495: parameter 0.3.  The resulting distributions of $\tau_o$ are so similar that it would be extremely difficult to distinguish between these eccentricity distributions.

1496: %

1497: {\em Bottom:} Same as above, but for histograms of $\tau_b$, the ratio

1498: of the observed transit duration to that expected for a similar

1499: planet, star, impact parameter, and orbital period, but a circular

1500: orbit.  Again, the distributions of $\tau_b$ are too similar to distinguish between the eccentricity distributions.

1501: %

1502: \label{FigDistribComp}

1503: }

1504: \end{figure}

1505: %

1506: %

1507: 

1508: 

1509: 

1510: \end{document}

1511: