1: \newcommand{\be}{\begin{equation}}
2: \newcommand{\ee}{\end{equation}}
3: \newcommand{\rstar}{\ensuremath{R_\star}}
4: \newcommand{\mstar}{\ensuremath{M_\star}}
5: \newcommand{\rpl}{\ensuremath{R_{p}}}
6: \newcommand{\mpl}{\ensuremath{M_{p}}}
7: \newcommand{\dotw}{\ensuremath{\dot{\omega}_{\rm GR}}\ }
8: %\documentclass[12pt,preprint]{aastex}
9: \documentclass[onecolumn,apj]{emulateapj}
10:
11:
12: \slugcomment{Accepted for publication in ApJ}
13: \lefthead{JORD\'AN \& BAKOS}
14: \righthead{GR PRECESSION IN EXOPLANETS}
15:
16: \begin{document}
17:
18: \title{
19: Observability of the General Relativistic Precession of Periastra in
20: Exoplanets
21: }
22:
23: \author{
24: Andr\'es Jord\'an\altaffilmark{1,2,3} and
25: G\'asp\'ar \'A. Bakos\altaffilmark{1,4}
26: }
27:
28: \altaffiltext{1}{
29: Harvard-Smithsonian Center for Astrophysics,
30: 60 Garden St., Cambridge, MA 02138;
31: ajordan@cfa.harvard.edu, gbakos@cfa.harvard.edu.
32: }
33: \altaffiltext{2}{Clay Fellow.}
34: \altaffiltext{3}{Departamento de Astronom\'{\i}a y Astrof\'{\i}sica,
35: Pontificia Universidad Cat\'olica de Chile, Casilla 306, Santiago 22,
36: Chile.}
37: \altaffiltext{4}{NSF Postdoctoral Fellow.}
38:
39:
40: \begin{abstract}
41: The general relativistic precession rate of periastra in close-in
42: exoplanets can be orders of magnitude larger than the magnitude of the
43: same effect for Mercury. The realization that some of the close-in
44: exoplanets have significant eccentricities raises the possibility that
45: this precession might be detectable. We explore in this work the
46: observability of the periastra precession using radial velocity and
47: transit light curve observations. Our analysis is independent of the
48: source of precession, which can also have significant contributions
49: due to additional planets and tidal deformations.
50: %
51: We find that precession of the periastra of the magnitude expected
52: from general relativity can be detectable in timescales of $\lesssim
53: 10$ years with current observational capabilities by measuring the
54: change in the primary transit duration or in the time difference
55: between primary and secondary transits. Radial velocity curves alone
56: would be able to detect this precession for super-massive, close-in
57: exoplanets orbiting inactive stars if they have $\sim 100$ datapoints
58: at each of two epochs separated by $\sim 20$ years.
59: %
60: We show that the contribution to the precession by tidal deformations
61: may dominate the total precession in cases where the relativistic
62: precession is detectable.
63: %
64: Studies of transit durations with {\it Kepler} might need to take into
65: account effects arising from the general relativistic and tidal induced
66: precession of periastra for systems containing close-in, eccentric
67: exoplanets. Such studies may be able to detect additional planets with
68: masses comparable to that of Earth by detecting secular variations in
69: the transit duration induced by the changing longitude of periastron.
70: %
71: \end{abstract}
72:
73: \keywords{celestial mechanics --- planetary systems}
74:
75:
76: \section{Introduction}
77:
78: Following the discovery of an extra-solar planet around the solar type
79: star 51 Pegasi \citep{Mayor1995a} there has been rapid progress in the
80: detection and characterization of extra-solar planetary systems.
81: %
82: The very early discoveries have shattered our view on planetary
83: systems, as certain systems exhibited short periods (51 Peg), high
84: eccentricities \citep[e.g.\ 70 Virginis b,][]{Marcy1996}, and massive
85: planetary companions \citep[e.g.\ Tau Boo b,][]{Butler1997a}.
86:
87: Interestingly, systems with all these properties combined
88: (i.e.~massive planets with short periods, small semi-major axes, high
89: eccentricities) have been also discovered
90: \citep[e.g., HAT-P-2b, XO-3b;][]{Bakos2007a,Johns-Krull2008a}. The high
91: eccentricities are somewhat surprising, as hot Jupiters with short
92: periods are generally expected to be circularized in timescales
93: shorter than the lifetime of the system if the parameter $Q$,
94: inversely proportional to the planet's tidal dissipation rate, is
95: assumed to be similar to that inferred for Jupiter
96: \citep{GoldreichSoter1966,Rasio1996}.
97:
98: By virtue of their small semi-major axes and high eccentricities, the
99: longitude of periastron $\omega$ of some of the newly discovered
100: systems are expected to precess due to General Relativistic (GR)
101: effects at rates of degrees per century. This is orders of magnitude
102: larger than the same effect observed in Mercury in our Solar System
103: ($43\arcsec$/century), which offered one of the cornerstone tests of
104: GR.
105: %
106: Furthermore, the massive, close-in eccentric planets induce
107: significant reflex motion of the host star, therefore enhancing the
108: detectability of the precession directly via radial velocities.
109:
110: In this work we explore the observability of the precession of the
111: longitude of periastron with the magnitude expected from GR in
112: exoplanets using radial velocity and transit timing observations. We
113: also consider in this work the periastra precession due to planetary
114: perturbers and tidal deformations, which can have contributions
115: comparable or greater than that of GR.
116: %
117: Previous works \citep{Miralda-Escude2002a,Heyl2007a} have explored
118: some aspects of the work presented here in the context of using timing
119: observations to detect terrestrial mass planets. We refer the reader
120: to independent work by \cite{Pal2008b} that also explores the
121: measurable effects of the periastra precession induced by GR.
122:
123: \section{Expected Precession of Periastra}
124: \label{sec:prec}
125:
126: Before continuing let us fix our notation. In what follows $a$ will
127: denote the semi-major axis of the Keplerian orbit of the planet-star
128: separation, $e$ its eccentricity, $P$ its
129: %
130: %
131: period, $\omega$ its longitude of periastron, $\mstar$ and $\rstar$
132: the mass and radius of the host star, respectively, and $n\equiv
133: (GM_{tot}/a^3)^{1/2}$ is the Keplerian mean motion (orbital angular
134: frequency), where $G$ is Newton's gravitational constant and $M_{\rm
135: tot}=\mstar+\mpl$ with $\mpl$ the mass of the planet. The reflex
136: motion of the host star is characterized in a similar manner. In this
137: section we detail the expected mechanisms that will cause a precession
138: in the value of $\omega$.
139:
140: \subsection{General Relativistic Precession}
141:
142: One of the most well-known consequences of General Relativity is that
143: orbits in a Schwarzschild metric are no longer closed as is the case
144: for the Kepler problem in Newtonian mechanics. The rate of precession
145: of the longitude of periastron due to GR is given to leading
146: post-Newtonian order by
147: %
148: \be
149: \dot{\omega}_{\rm GR} = \frac{3 G \mstar }{a c^2 (1-e^2)}n
150: \ee
151: \citep[see any general relativity textbook, e.g.][for a
152: derivation]{MTW}. With $a$ expressed in astronomical units, $P$ in
153: days and the mass of the host star $M_*$ in solar units,
154: $\dot{\omega}_{\rm GR}$ is given in units of degrees per century by
155: \be
156: \dot{\omega}_{\rm GR} = \frac{7.78}{(1-e^2)}
157: \left(\frac{\mstar}{M_\odot}\right)
158: \left(\frac{a}{0.05\mbox{AU}}\right)^{-1}
159: \left(\frac{P}{\mbox{day}}\right)^{-1}
160: \,\,\,\,[^\circ/\mbox{century}].
161: \label{eq:wdotGR}
162: \ee
163: We use equation~\ref{eq:wdotGR} to estimate the expected precession
164: for currently known exoplanets as listed in the online California \&
165: Carnegie Catalog of Exoplanets \citep[][version Nov 7
166: 2007]{ButlerCat2006}. We list in Table~\ref{tab:prec} all exoplanets
167: that have $e>0.1$ and that have $\dot{\omega}_{\rm GR} > 1
168: ^\circ/\mbox{century}$\footnote{The magnitude of all the effects
169: discussed in what follows where \dotw would manifest itself are increasing
170: functions of $e$. Systems with low $e$ are therefore not relevant
171: from the point of view of detecting GR effects. \citet{Heyl2007a}
172: presents in their Figure~4 an estimate of \dotw for all systems in
173: \citet{ButlerCat2006} and note the four systems with higher
174: values. None of those are in Table~\ref{tab:prec} because they all
175: have $e< 0.03$.}. From left to right, the columns in
176: Table~\ref{tab:prec} record the name of the exoplanet, its semi-major
177: axis, its eccentricity, the mass $\mstar$ of the host star, the
178: period, the velocity amplitude\footnote{This is often referred to as the
179: semi-amplitude in the exoplanet literature. We choose to use simply
180: amplitude in this paper in order to agree with the widespread usage in
181: the physical sciences for the multiplicative factor in a simple
182: harmonic oscillator.}, the longitude of periastron $\omega$, the
183: calculated $\dot{\omega}_{\rm GR} $ and finally a flag that is 1 if
184: the planet is transiting its host star and 0 otherwise.
185: %
186: The condition on $e$ is in order to consider only systems where GR
187: effects can be constrained with sufficient confidence. Additionally,
188: we restrict ourselves to stars which have evidence currently for a
189: single exoplanet in order to avoid the complications arising from the
190: precession of $\omega$ induced by other planets (see
191: below)\footnote{\citet{Adams2006a,Adams2006b} studied the effects of
192: secular interactions in multiple-planet systems including the effects
193: of GR. They show that GR can have significant effects on secular
194: perturbations for systems with favorable characteristics.}. We have
195: added to the exoplanets listed by \citet{ButlerCat2006} the recently
196: discovered XO-3b
197: \citep{Johns-Krull2008a}, which has the largest predicted
198: $\dot{\omega}_{\rm GR}$ of all exoplanets listed in Table~\ref{tab:prec}.
199:
200: \begin{deluxetable}{lccccrrcc}
201: \tablecaption{GR precession of exoplanets with $e>0.1$ and GR
202: Precession Rates $>1^\circ$/century}
203: \tabletypesize{\footnotesize}
204: \tablewidth{0pt}
205: \tablehead{
206: \colhead{Name} & \colhead{$a$} & \colhead{$e$}
207: & \colhead{$\mstar$ } & \colhead{Period}
208: & \colhead{K} & \colhead{$\omega$}
209: & \colhead{$\dot{\omega}_{\rm GR} $} & \colhead{TEP?}\\
210: \colhead{} & \colhead{(AU)} & \colhead{}
211: & \colhead{$(M_\odot)$} & \colhead{(days)}
212: & \colhead{(m sec$^{-1}$)} & \colhead{(deg)}
213: & \colhead{$^\circ$/century} & \colhead{}
214: }
215: \startdata
216: HD49674 b & 0.058 & 0.290 & 1.060 & 4.944 & 13.7 & 283.0 & 1.576 & 0 \\
217: HD88133 b & 0.047 & 0.133 & 1.200 & 3.416 & 36.1 & 349.0 & 2.958 & 0 \\
218: GJ 436 b & 0.028 & 0.159 & 0.410 & 2.644 & 18.7 & 339.0 & 2.234 & 1 \\
219: HD118203 b & 0.070 & 0.309 & 1.230 & 6.133 & 217.0 & 155.7 & 1.231 & 0 \\
220: HAT-P-2 b & 0.069 & 0.507 & 1.350 & 5.633 & 884.0 & 184.6 & 1.836 & 1 \\
221: HD185269 b & 0.077 & 0.296 & 1.280 & 6.838 & 90.7 & 172.0 & 1.046 & 0 \\
222: XO-3 b & 0.048 & 0.260 & 1.410 & 3.192 & 1471.0 & -15.4 & 3.886 & 1 \\
223: \enddata
224: \label{tab:prec}
225: \end{deluxetable}
226:
227:
228: As can be seen in Table~\ref{tab:prec}, seven systems have
229: $\dot{\omega}_{\rm GR} > 1 ^\circ/\mbox{century}$, with 3 of them
230: being transiting systems. In timescales of a few tens of years, the
231: longitude of periastron of the systems is expected to shift in these
232: systems by $\delta \omega \gtrsim 0.5^\circ$, a change that, as we
233: will show below, may produce detectable effects. GR effects are not
234: the only mechanisms that can cause a shift in $\omega$ though, so we
235: now turn our attention to additional mechanisms.
236:
237: \subsection{Stellar Quadrupole, Tides and Additional Planets}
238: \label{ssec:add_prec}
239:
240: Besides the GR precession discussed above, $\omega$ can precess due to
241: the presence of additional effects. \citet{Miralda-Escude2002a}
242: discussed the observability, using the duration of transits and the
243: period between transits, of changes in $\omega$ caused by a stellar
244: quadrupole moment and perturbations from other planets. These effects,
245: some of which were discussed using more accurate calculations by
246: \citet{Heyl2007a}, may additionally cause a precession of the orbital plane.
247: %
248: Additionally, tidal deformations induced on the star and the planet
249: can also produce a secular change on $\omega$, an effect not
250: considered in the studies mentioned above. The effect of apsidal
251: motions induced by tidal deformations is a well known effect in
252: eclipsing binaries
253: \citep{Sterne1939a,Quataert1996a,Smeyers2001a}, and was included in
254: the analysis of the planetary system around HD~83443 by
255: \citet{Wu2002a}. Tidal deformations can produce a significant amount
256: of precession in close-in exoplanets and should therefore be taken
257: into account.
258:
259: The precession caused by a stellar quadrupole moment is given
260: to second order in $e$ and first order in $(R_*/a)^2$ by
261: \be
262: \dot{\omega}_{\rm quad} \approx \frac{3 J_2 \rstar^2}{2a^2} n,
263: \ee
264: \noindent where $J_2$ is the quadrupole moment
265: \citep{MurrayDermott}. In units of degree/century this expression
266: reads
267: \be
268: \dot{\omega}_{\rm quad} \approx 0.17
269: \left(\frac{P}{\mbox{day}}\right)^{-1}
270: \left(\frac{J_2}{10^{-6}}\right)
271: \left(\frac{\rstar}{R_\odot}\right)^2
272: \left(\frac{a}{0.05\mbox{AU}}\right)^{-2}
273: \,\,\,\,[^\circ/\mbox{century}].
274: \ee
275: It is clear from this equation that for values of $J_2\lesssim
276: 10^{-6}$ similar to that of the Sun \citep{Pireaux2003a} the value of
277: $\dot{\omega}_{\rm quad}$ is smaller than the value of
278: $\dot{\omega}$ expected from GR \citep[see
279: also][]{Miralda-Escude2002a}. We will therefore assume in what follows
280: that $\dot{\omega}_{\rm quad}$ is always negligible in comparison with
281: $\dot{\omega}_{\rm GR}$.
282:
283: The tidal deformations induced on the star and the planet by each
284: other will lead to a change in the longitude of periastron which is
285: given, under the approximation that the objects can instantaneously
286: adjust their equilibrium shapes to the tidal force and considering
287: up to second order harmonic distortions, by
288: %
289: \be
290: \dot{\omega}_{\rm tide} \approx \frac{15 f(e)}{a^5}
291: \left( \frac{k_{2,s}\mpl \rstar^5}{\mstar} +
292: \frac{k_{2,p}\mstar \rpl^5}{\mpl}
293: \right) n,
294: \label{eq:tideN}
295: \ee
296: \noindent where $f(e)\equiv(1-e^2)^{-5}[1+(3/2)e^2+(1/8)e^4]$,
297: and $k_{2,s}, k_{2,p}$ are the apsidal motion constants for the star
298: and planet respectively, which depend on the mass concentration of the
299: tidally deformed bodies \citep{Sterne1939a}. For stars we expect
300: $k_{2,s} \lesssim 0.01$ \citep{Claret1992a}, while for giant planets
301: we expect $k_{2,p} \approx 0.25$ if we assume that their structure can
302: be roughly described by a polytrope of index $n
303: \approx 1$ \citep{Hubbard1984a}. For the extreme case of a sphere of uniform mass
304: density, the apsidal motion constant takes the value $k_2=0.75$
305: \citep[e.g.,][]{Smeyers2001a}.
306: %
307: We see from Equation~\ref{eq:tideN} that for close-in hot Jupiters the
308: term containing $k_{2,p}$ will dominate, and that the effect of tides
309: on $\omega$ increases very rapidly with decreasing $a$.
310: In units of degree/century equation~\ref{eq:tideN} gives
311: \be
312: \dot{\omega}_{\rm tide} \approx
313: 1.6 f(e) {\cal T}
314: \left(\frac{P}{\mbox{day}}\right)^{-1}
315: \left(\frac{k_{2,p}}{0.1}\right)
316: \left(\frac{a}{0.05\mbox{AU}}\right)^{-5}
317: \left(\frac{\rpl}{R_J}\right)^5
318: \left(\frac{M_J}{\mpl}\right)
319: \left(\frac{\mstar}{M_\odot}\right)
320: \,\,\,\,[^\circ/\mbox{century}],
321: \ee
322: \noindent where we have introduced
323: ${\cal T} \equiv 1 + (\rstar/\rpl)^5(\mpl/\mstar)^2(k_{2,s}/k_{2,p})$, which
324: is $\approx 1$ for the case of a close-in Jupiter.
325: %
326: Assuming that $k_{2,p} \sim 0.1$, $e\lesssim 0.5$, $\mpl \sim M_{J}$,
327: $\mstar \sim M_\odot$, $\rstar \sim R_\odot$ and $\rpl \sim R_J$
328: it follows that $\dot{\omega}_{\rm tide}$ is of comparable
329: magnitude as $\dot{\omega}_{\rm GR}$.
330:
331: The precession of the periastra caused by a second planet, which we
332: dub a ``perturber'', is given to first order in $e$ and lowest order
333: in $(a/a_2)$ by
334: \be
335: \dot{\omega}_{\rm perturber} \approx \frac{3M_2 a^3}{4 \mstar a_2^3}n
336: \label{eq:wdot_pert}
337: \ee
338: \citep{MurrayDermott,Miralda-Escude2002a}, where $M_2$ is the
339: mass of the second planet and $a_2$ the semi-major axis of its
340: orbit. In terms of deg/century this expression reads
341: \be
342: \dot{\omega}_{\rm perturber} \approx 29.6
343: \left(\frac{P}{\mbox{day}}\right)^{-1}
344: \left(\frac{a}{a_2}\right)^{3}
345: \left(\frac{\mstar}{M_\odot}\right)^{-1}
346: \left(\frac{M_2}{M_\oplus}\right)
347: \,\,\,\,[^\circ/\mbox{century}].
348: \ee
349:
350: For a perturber with $a_2 = 2a$ and and a mass similar to Earth
351: orbiting a solar-mass star, we get that $\dot{\omega}_{\rm perturber}
352: \sim 3\times 10^{-7}n$ or $\dot{\omega}_{\rm perturber}
353: \sim 0.7 $ deg/century for a $P=5$ days planet. This is comparable
354: to the precession expected from GR and therefore any detected
355: precession of the longitude of periastron will be that of GR plus the
356: possible addition of any perturber planet present in the system and
357: the effects of tidal deformations (and generally a negligible
358: contribution from the stellar quadrupole).
359:
360: In what follows we will discuss the observability of changes in
361: $\omega$ using radial velocity and transit observations. As just
362: shown, any precession is expected to arise by GR, the effect of
363: additional planets or tidal deformations. The discussion that follows
364: addresses the detectability of changes in $\omega$ independent of
365: their origin.
366:
367: \section{Observability of Periastra Precession in Extra-Solar Planets}
368:
369: \subsection{Radial Velocities}
370: \label{subsec:rv}
371:
372: The radial velocity of a star including the reflex motion due to a
373: planetary component is given by
374: \be
375: v_r(t) = v_0 + K[\cos(\omega + f(t-t_0)) + e\cos(\omega)],
376: \label{eq:rv}
377: \ee
378: where $v_0$ is the systemic velocity, $t_0$ the time
379: coordinate zeropoint, $f$ the true anomaly and $K$ is the velocity
380: amplitude which is related to the orbital elements and the masses
381: by
382: \be
383: K = \left( \frac{2\pi G}{P}\right)^{1/3} \frac{\mpl \sin i}{M_{\rm tot}},
384: \ee
385:
386: \noindent where $i$ is the orbit inclination. Fitting for the observed
387: radial velocities of a star will give then direct estimates of $v_0,
388: t_0, K, e, P$ and $\omega$.
389:
390: Our aim in this section is to determine if $\omega$ can be constrained
391: tightly enough in timescales of tens of years or less in order to
392: detect changes in $\omega$ of the magnitude produced by GR in those
393: time-spans.
394: %
395: In order to do this we have simulated data and then fit it with a
396: model of the form given by equation~\ref{eq:rv} a total of 1000
397: times. We then recover the best-fit values of $\omega$ in all
398: simulations and use that to estimate the probability distribution
399: $\phi(\omega)$ expected under given assumptions.
400:
401: The systemic velocity $v_0$ and $t_0$ are just zero-points that we set
402: to 0 in all our simulations, where we also set the time units such
403: that $P=1$ (note though that we do fit for all these quantities so
404: that their effect on the fit propagates to the uncertainties of
405: $\omega$). By trying several values of $\omega$ we have verified that
406: the probability distributions recovered do not depend strongly on the
407: particular value of $\omega$, which we therefore fix for all
408: simulations at an arbitrary value $\omega_0=135^\circ$. This leaves
409: us with just two parameters to vary, namely $e$ and $K$.
410:
411: Given $e$ and $K$ we generate $N_{\rm obs}$ data-points with times
412: $t_i$ uniformly distributed\footnote{We ignore in
413: our simulations the Rossiter-McLaughlin effect for the case of
414: transiting planets
415: \citep{Rossiter1924a,McLaughlin1924a,Queloz2000a}.}
416: within a period and then we add to each
417: time a random number of periods between 0 and 20. For transiting
418: exoplanets the observations would have to be taken uniformly in time
419: intervals excluding the transit. We then get the observed radial
420: velocity from equation~\ref{eq:rv} as
421:
422: \be
423: v_{r}(t_i) = K[\cos(\omega_0 + f(t_i)) + e\cos(\omega_0)] +
424: G(0,\sigma_{\rm tot})
425: \label{eq:vr}
426: \ee
427:
428: \noindent where $G(0,\sigma_{\rm tot})$ is a random Gaussian deviate
429: with mean 0 and standard deviation $\sigma_{\rm tot}$. The latter
430: quantity is obtained as $\sigma_{\rm tot}^2 = \sigma_{\rm obs}^2 +
431: \sigma_{\rm jitter}^2$, where $\sigma_{\rm obs}$ is the random
432: uncertainty for each measurement and $\sigma_{\rm jitter}$ is the
433: noise arising from stellar jitter. Given the form of
434: equation~\ref{eq:rv} a Fisher matrix analysis implies that the
435: uncertainty in the longitude of periastron, $\sigma_\omega$ satisfies
436: the following scaling
437:
438: \be
439: \sigma_\omega \propto N_{\rm obs}^{-1/2}\sigma_{\rm tot} K^{-1}.
440: \label{eq:scaling}
441: \ee
442:
443:
444: This scaling allows us to perform a set of fiducial simulations for
445: several values of $e$ and use the results to scale to parameters
446: relevant to a given situation of interest.
447: %
448: We note that the simulations we performed verify that the scaling
449: inferred from a Fisher matrix analysis is accurate.
450: %
451: In order to measure the longitude of periastron $\omega$ with a
452: reasonable degree of certainty the system clearly needs to have a
453: significant amount of eccentricity. We restrict ourselves to systems
454: with $e\ge0.1$ and we simulate systems with
455: $e=0.1,0.2,0.3,0.4,0.5,0.6$.
456:
457: For $\sigma_{\rm obs}$ we assume a typical high-precision measurement
458: with $\sigma_{\rm obs}=2$ m sec$^{-1}$. For the stellar jitter, we
459: perform our fiducial simulations for a typical jitter of $\sigma_{\rm
460: jitter} = 4$ m sec$^{-1}$ \citep{wright2005,ButlerCat2006}.
461: %
462: The limitations imposed by active stars and/or different precision on
463: the radial velocity measurements on the recovery of \dotw can be
464: explored by using the scaling of $\sigma_\omega$ with higher assumed
465: values of $\sigma_{\rm jitter}$ and/or $\sigma_{\rm obs}$.
466: %
467: Even though Equation~\ref{eq:scaling} renders multiple values of $K$
468: redundant, we choose to present results for two values of $K$ for
469: illustrative purposes, namely $K=100$ m sec$^{-1}$ and $K=1000$ m
470: sec$^{-1}$. The former corresponds roughly to Jupiter-mass exoplanets
471: and is fairly representative of currently known systems
472: \citep{ButlerCat2006}\footnote{See also \url{http://www.exoplanet.eu}.},
473: while the latter corresponds to super-massive planets which as we will
474: see are the class of systems which would allow the detection of
475: $\dot{\omega}_{\rm GR}$ with radial velocities.
476: %
477: Finally, we do our fiducial simulations for $N_{\rm obs} =100$, a
478: value not atypical for well-sampled radial velocity curves available
479: today.
480:
481: \begin{deluxetable}{cccc}
482: \tablecaption{Results of Radial Velocity Curves Fit Simulations for
483: $N_{\rm obs}=100$, $\sigma_{\rm obs}=2$ m/sec and
484: $\sigma_{\rm jitter}=4$ m/sec}
485: \tabletypesize{\footnotesize}
486: \tablewidth{0pt}
487: \tablehead{
488: \colhead{$K$} &
489: \colhead{$e$ } &
490: \colhead{$\sigma_\omega$} &
491: \colhead{$\alpha_{20}$} \\
492: \colhead{(m sec$^{-1}$)}
493: & \colhead{} &
494: \colhead{(deg)} &
495: \colhead{(deg/century)}
496: }
497: \startdata
498: %\input{tab2}
499: 100 & 0.10 & 3.77 & 80.010\\
500: 100 & 0.20 & 1.85 & 39.204\\
501: 100 & 0.30 & 1.29 & 27.457\\
502: 100 & 0.40 & 1.05 & 22.249\\
503: 100 & 0.50 & 0.95 & 20.253\\
504: 100 & 0.60 & 0.86 & 18.159\\
505: 1000 & 0.10 & 0.38 & 8.059\\
506: 1000 & 0.20 & 0.19 & 4.106\\
507: 1000 & 0.30 & 0.13 & 2.768\\
508: 1000 & 0.40 & 0.10 & 2.182\\
509: 1000 & 0.50 & 0.09 & 1.959\\
510: 1000 & 0.60 & 0.08 & 1.790\\
511: \enddata
512: \label{tab:sim}
513: \end{deluxetable}
514:
515:
516: The results of the simulations are summarized in Table~\ref{tab:sim}.
517: From left to right, the columns in this Table record the assumed
518: radial velocity amplitude $K$, the system eccentricity $e$, the
519: expected uncertainty in the longitude of periastron $\sigma_\omega$
520: for the simulated system, and finally $\alpha_{20}$, which we define
521: to be the value of $\dot{\omega}$ that would be necessary in order to
522: achieve a 3$\sigma$ detection in the simulated systems in a time-span
523: of 20 years when measuring $\omega$ in two epochs, each having $N_{\rm
524: obs}$ observations. We note that we have performed Shapiro-Wilk
525: normality tests in the recovered $\omega$ distributions. We found that
526: all of them are consistent with normality and thus we are justified in
527: using the dispersion $\sigma_w$ to derive confidence levels.
528:
529:
530: \begin{figure}
531: \epsscale{0.6}
532: \plotone{f1.eps}
533: \caption{
534: Distribution $\phi$ of recovered angles of periastron
535: $\omega-\omega_0$, where $\omega_0$ is the input angle, for the
536: simulation with $N_{\rm obs}=100$, $K=1000$ m sec$^{-1}$ and
537: $\sigma_j=4$ m sec$^{-1}$. The distributions are shown for
538: eccentricities $e=0.1,0.3,0.5$.
539: \label{fig:rv_eg}
540: \vspace{0.2cm}
541: }
542: \end{figure}
543:
544: As an example, we show in Figure~\ref{fig:rv_eg} the distributions of
545: recovered $\omega$ for the case with $N_{\rm obs}=100$, $K=1000$ m
546: sec$^{-1}$ and $\sigma_j=4$ m sec$^{-1}$ for eccentricities
547: $e=0.1,0.3,0.5$.
548: %
549: In the upper panel of Figure~\ref{fig:rv_eg2} we show the radial
550: velocity curve for a system with $K=1000$ m sec$^{-1}$, $e=0.5$ and
551: $\omega=135^\circ$, while in the lower panel we show the difference
552: between that curve and a radial velocity curve having
553: $\omega=136^\circ$. The simulations presented in Table~\ref{tab:sim}
554: show that this difference can be detected at the 10-$\sigma$ level
555: ($\sigma_\omega \sim 0.1^\circ$) when using $N_{\rm obs}=100$
556: observations, each having an uncertainty of $\sim 2$ m sec$^{-1}$.
557: The high level of significance can be achieved in this case thanks to
558: the assumed low-level of jitter and the super-massive nature of a
559: system with that semi-velocity amplitude.
560:
561: \begin{figure}
562: \epsscale{0.6}
563: \plotone{f2.eps}
564: \caption{
565: ({\it Top}\/) Radial velocity curve as a function of orbital phase
566: for a system with $K=1000$ m sec$^{-1}$, $e=0.5$ and
567: $\omega=135^\circ$. ({\it Bottom}\/) Difference between the curve in
568: the top panel and a radial velocity curve that is identical except
569: in that it has $\omega=136^\circ$. Note the different $y$-axis
570: scales used in the different panels.
571: \label{fig:rv_eg2}
572: \vspace{0.2cm}
573: }
574: \end{figure}
575:
576: In order to validate our simulations against uncertainty estimates
577: obtained with real data, we have run a simulation with $N_{\rm
578: obs}=20$, $e=0.517$, $K=1011$ m sec$^{-1}$, $\omega=179.3^\circ$ and $\sigma_j=60$ m
579: sec$^{-1}$. These observational conditions and orbital parameters are
580: appropriate for the observations of HAT-P-2 b reported by
581: \citet{Bakos2007a}. Our simulations for this case return
582: $\sigma_w=3.8^\circ$, while \citet{Bakos2007a} report $\omega_{\rm
583: HAT-P-2 b} = 179.3 \pm 3.6^\circ$, in very good agreement with our
584: estimate. We conclude that our simulations return realistic estimates
585: of the uncertainties in $\omega$.
586:
587: The simulations presented in Table~\ref{tab:sim} show that in some
588: cases the precession of periastron detectable in 20 years,
589: $\alpha_{20}$, is comparable to the values $\dot{\omega}_{GR}$ of
590: currently known systems listed in Table~\ref{tab:prec}. For close-in,
591: eccentric, super-massive planets ($K\sim 1000$ m sec$^{-1}$) about 100
592: observations per epoch are sufficient, while for Jupiter-mass
593: exoplanets ($K\sim100$ m sec$^{-1}$) an unrealistically large number
594: of radial velocity observations, $N_{\rm obs}=10000$ per epoch, would
595: be needed. Therefore, observations of different epochs of radial
596: velocities with a time-span of $\sim$ 20 years would detect the
597: variations in the precession of exoplanet periastra induced by GR in
598: some currently known super-masive systems. The estimates above assume
599: a typical level of stellar jitter. For active stars the number of
600: observations have to be increased in proportion to the dispersion
601: characterizing the jitter (see Equation~\ref{eq:scaling}). Stellar
602: activity is therefore an important limitation to detect changes in
603: $\omega$ using radial velocities. All in all, radial velocity studies
604: of exoplanets will be generally able to ignore the effects of GR
605: precession as they will usually be well below a detectable level.
606:
607: We note in closing that \citet{Miralda-Escude2002a} and
608: \citet{Heyl2007a} consider using radial velocities and transit timing
609: observations
610: in order to measure the small difference between the period observed
611: in the radial velocities and the period between primary transits.
612: Both works conclude that this is not a competitive method and so we
613: will not consider it further here and refer the reader to those works
614: for details.
615:
616: \subsection{Duration of Transits}
617: \label{subsec:D}
618:
619: As the planetary orbit acquires a significant eccentricity $e$, the
620: duration of the primary and secondary transits are no longer equal and
621: acquire a dependence on the longitude of periastron of the system. An
622: explicit expression for the duration of a transit $D$ in the eccentric
623: case was derived by \citet[][their equation 7]{Tingley2005} under the
624: assumption that the distance between the planet and the star does not
625: change significantly during transit. It is given by
626: \be
627: D = 2Z(\rstar + \rpl)\frac{\sqrt{1-e^2}}{(1+e\cos(f_t))}
628: \left(\frac{P}{2\pi G M_{\rm tot}}\right)^{1/3},
629: \ee
630:
631: \noindent where
632:
633: \be
634: Z = \sqrt{1-\frac{r_t^2\cos^2 i}{(\rstar+\rpl)^2}} \equiv \sqrt{1-b^2}
635: \ee
636:
637: \noindent is a geometrical factor related to the impact parameter
638: $b$. $\rstar$ and $\rpl$ are the radii of the star and planet
639: respectively, $M_{\rm tot} \equiv \mstar + \mpl$ and $r_t$ and $f_t$
640: are the radii and true anomaly at the time of transit,
641: respectively. The latter clearly depends linearly on $\omega$ and we
642: therefore have $\dot{f}_t = \dot{\omega}$. The logarithmic derivative
643: of the duration of a transit is
644:
645: \be
646: d\ln D/dt = \frac{\dot{\omega} e \sin(f_t)}{(1+e\cos(f_t))}
647: \left\{1-\frac{b^2}{1-b^2}\right\}.
648: \label{eq:dlnD}
649: \ee
650:
651: In Figure~\ref{fig:dlnDdt} we show the quantity $(1/\dot{\omega}) d\ln
652: D/dt$ for a central transit ($b=0$) and for $e=0.1,0.3,0.5,0.7$, with
653: higher $e$ giving higher values of $|(1/\dot{\omega}) d\ln D/dt|$. The
654: change in the duration of an eclipse for a small change in $\omega$
655: given by $\dot{\omega}\delta t$ is simply $\delta D \sim D
656: \dot{\omega}\delta t \,\,[\dot{\omega}^{-1}d\ln D/dt] $. For
657: $\dot{\omega}\delta t \sim 0.5\times10^{-2}$ rad, appropriate for the
658: expected change in $\omega$ for $\dotw \sim 3$ deg/century over 10
659: years, we have that $\delta D \sim 0.075D \times 10^{-2}$ for $e \sim
660: 0.3$, which translates into $\delta D\sim 10$ sec for a transit
661: duration of $D \sim 0.15$ day which is typical for the sytems that we
662: explore in this work.
663:
664: \begin{figure}
665: \epsscale{0.6}
666: \plotone{f3.eps}
667: \caption{
668: $(1/\dot{\omega}) d\ln D/dt$ as a function of the true anomaly
669: at the time of transit $f_t$ for a central transit
670: ($p=0$). The different curves are for difference
671: eccentricities $e=0.1,0.3,0.5,0.7$, with higher $e$ giving
672: higher values of $|(1/\dot{\omega}) d\ln D/dt|$
673: \label{fig:dlnDdt}. The extrema in this figure are at values
674: of the true anomaly at the time of transit of $f_t=
675: \pm\arccos(-e)$. }
676: \end{figure}
677:
678:
679:
680: A very interesting feature of Equation~\ref{eq:dlnD} is its dependence
681: on the impact parameter $b$. First, $d\ln D/dt$ vanishes for
682: $b=1/\sqrt{2}$. This behavior is possible due to two competing effects
683: which cancel out exactly for that value of $b$: an increase/decrease
684: in the star-planet separation causes both and decrease/increase in the
685: path-length of the planet across the disk of the star and a
686: corresponding decrease/increase on the velocity across it, which
687: implies a slower/faster crossing-time. Secondly, for systems with
688: values of $b$ close to 1 (i.e., near-grazing systems), the value of
689: $d\ln D/dt$ increases greatly. We will come back to near-grazing
690: systems below; in what immediately follows we will quantify the
691: accuracy to which we can determine the transit duration $D$.
692:
693:
694: If the times of beginning of ingress and end of egress are denoted by
695: $t_i$ and $t_e$ respectively, then the duration of a transit is given
696: by $D=t_e-t_i$ and the uncertainty in the duration is
697: $\sigma_D^2=\sigma_{t_i}^2 + \sigma_{t_e}^2$ (assuming no correlation
698: between $t_i$ and $t_e$). If the ingress has a
699: duration of $\Delta t_i$ (i.e.~the time from first contact to second
700: contact) then a linear approximation of the flux during this time can
701: be written as $F(t) =
702: F_0(1-(t-t_i)(\rpl/\rstar)^2/\Delta t_i)$, where $F_0$ is the
703: out-of-transit stellar flux and we ignore the effects of limb
704: darkening. If the light curve is being sampled at a rate $\Gamma$ per
705: unit time then we have $N= \Delta t_i \Gamma$ photometric measurements
706: during the egress. If each photometric measurement has a fractional
707: precision $\sigma_{\rm ph}$, and assuming $\rpl$ and $\rstar$ are
708: known, a least-squares fit to the photometric series will allow the
709: determination of $t_i$ with an uncertainty $\sigma_{t_i} = \sigma_{\rm
710: ph} \Delta t_i (\rstar/\rpl)^2 N^{-1/2} =
711: \sigma_{\rm ph} (\Delta t_i/\Gamma)^{1/2}(\rstar/\rpl)^2$.
712: The uncertainty in the time of egress is given by a similar expression
713: replacing $\Delta t_i$ by $\Delta t_e$.
714:
715: If we assume that the ingress and egress times are equal\footnote{The
716: ingress and egress times are not equal in general for eccentric
717: systems, see Equation~7 in \citet{Ford2008a}. This small difference
718: has no significant effect on the uncertainty estimates dealt with
719: here.}, and insert explicit expressions for $\Delta t_i \approx
720: \Delta t_e$ and $D$, we get that for a central transit
721: \citep{Ford2008a}:
722:
723: %\begin{widetext}
724: \be
725: \frac{\sigma_D}{D} \approx \frac{40.7 \sigma_{\rm ph}\sqrt{1+e\cos f_t}}{(1-e^2)^{1/4}}
726: \sqrt{\frac{2}{N_{tr}\Gamma}}
727: \left( \frac{\rpl}{R_\oplus}\right)^{-3/2}
728: \left( \frac{\rstar}{R_\odot}\right)
729: \left( \frac{\mstar}{M_\odot}\right)^{1/6}
730: \left( \frac{P}{{\rm yr}}\right)^{-1/6}
731: (1+\rpl/\rstar)^{-3/2},
732: \label{eq:sigmaD}
733: \ee
734: %\end{widetext}
735:
736: \noindent where $N_{tr}$ is the number of transits observed, $\Gamma$
737: is to be expressed in units of min$^{-1}$, and we have neglected
738: factors including $(1+\mu)$, where $\mu=\mpl/\mstar$. We note for later reference that the
739: uncertainty in the central transit time $t_c \equiv 0.5(t_e + t_i)$ is
740: $\sigma_ c \approx \sigma_D/\sqrt{2}$.
741:
742: The highest photometric precision in a transit light curve has been
743: achieved with {\it HST} \citep{Brown2001a,Pont2007a} with $\sigma_{\rm ph}
744: \sim 10^{-4}$. For parameters appropriate to HD~209458
745: Equation~\ref{eq:sigmaD} gives $\sigma_D \sim 5$ sec. The
746: uncertainties in the central time $\sigma_c \sim \sigma_D$ reported in
747: \citet{Brown2001a} are of the same order and we therefore deem
748: Equation~\ref{eq:sigmaD} to be a reasonable estimate of the expected
749: uncertainties\footnote{We note that \citet{Brown2001a} warn about the
750: presence of systematic effects not accounted for in the Poisson
751: uncertainty estimate that could be of the same order as $\sigma_D$ for
752: their observations.}. We will take this to be the highest precision
753: currently possible for facilities such as {\it HST} where only a few
754: transits are typically observed.
755:
756: The {\it Kepler} mission \citep{Borucki2003a,Basri2005a} will observe $\sim$100,000 stars
757: continuously with $\sigma_{\rm ph} \sim 4 \times 10^{-4}$ for a one
758: minute exposure of $V=12$ solar-like star and expects to find
759: a large number of close-in ``hot Jupiters''. The precision attainable in one
760: year for a 1-minute sampling of a Jupiter-mass system with $P=5$ days
761: orbiting a $V=12$ solar-type star is $\sigma_D/D \sim
762: 1.1\times10^{-4}$ or $\sigma_D \sim 1.5$ sec for a 0.15 days
763: transit duration.
764:
765: {\it Kepler} should therefore be capable of detecting changes in the
766: transit duration $D$ due to GR within its mission for some eccentric,
767: close-in systems, and certainly when coupled to follow-up
768: determinations of $D$ with a facility delivering a precision similar
769: to what {\it HST} can achieve.
770: %
771: Of particular interest will be systems that are near-grazing, due to
772: the factor of $(1-b^2)^{-1}$ in Equation~\ref{eq:dlnD}. If eccentric,
773: close-in, near-grazing systems are found in the {\it Kepler} CCDs,
774: they will be subject to significant changes in $D$ which should be
775: observable. For example, a system with $e=0.4$, $b=0.85$, $f_t=\pi/2$
776: and $\dotw \sim 3$ deg/century would have a change in $D$ of
777: $\sim 9$ sec during {\it Kepler's} lifetime for $D=0.075$ days, a
778: change which would be detectable. Of course, as $b$ approaches one, $D$
779: will tend to zero and the assumptions leading to
780: Equation~\ref{eq:sigmaD} will break down and make the estimated
781: uncertainty optimistic, both effects which will counter the
782: corresponding increase of $d\ln D/dt$.
783:
784:
785: \subsection{Period Between Transits}
786: \label{subsec:dPdt}
787:
788: As already noted by \citet{Miralda-Escude2002a} and \citet{Heyl2007a},
789: as the longitude of periastron changes, the period between transits
790: $P_t$ will change as well. Periods are the quantities that are
791: measured with the greatest precision, usually with uncertainties
792: on the order of seconds from ground-based observations
793: (e.g., the average uncertainty for HATNet planets discovered to date is
794: 4.5 seconds).
795:
796: To first order in $e$ the derivative of the transit period is given by
797:
798: \be
799: \dot{P}_t = 4\pi e\left(\frac{\dotw}{n}\right)^2\sin (M_t)
800: \ee
801:
802: \noindent where $M_t$ is the mean anomaly at transit
803: \citep{Miralda-Escude2002a} and is related to the true anomaly at
804: transit $f_t$ to first order in $e$ by $M_t = f_t - 2e\sin f_t$. For
805: $e=0.1$, a 5 day period, and $\dotw=3$ deg/century, the root mean
806: square value of $dP_t/dt$ over all possible $M_t$ values is $\sim
807: 10^{-12}$, which translates into a period change of $\sim 2 \times
808: 10^{-4}$ sec in 10 years. The period can be determined to a precision
809: $\sim \sigma_c N_{tr}^{-3/2}$ or $\sigma_D N_{tr}^{-1}2^{-1/2}$ using
810: the expression for $\sigma_D$ above\footnote{The scaling $N^{-3/2}$
811: follows from describing the central transit times as $t_i = t_0 + Pi$,
812: with $i=1, \ldots, N_{tr}$ and determining $P$ using a $\chi^2$
813: fit. The variance in the derived $P$ is \citep[see,
814: e.g.,][]{Gould2003a} $\sigma^2_P \approx (3/N^3)\sigma_c^2 + {\cal
815: O}(N^{-4})$.}.
816: %
817: Assuming the parameters for {\it Kepler} as above ($V=12$ star, $P=5$
818: day period) the precision achievable during 1 year is $\sim 0.013$
819: sec.
820:
821: Based on the numbers above we conclude that measuring significant
822: changes in the transiting period in $\lesssim 10 $ years timescales is
823: not feasible. Our conclusions are in broad agreement with the
824: analysis presented in \citet{Miralda-Escude2002a} and
825: \citet{Heyl2007a}, who conclude that thousands of transits need to be
826: observed with high precision in order to detect significant variations
827: in $P_t$. As there is no existing or planned facility that will allow
828: to observe this amount of transits with the required photometric
829: precision we conclude that measurements of $\dot{P}_t$ will not be
830: significantly affected by changes in $\omega$ of the magnitudes expected to
831: arise from GR or from the secular changes due to a perturber.
832:
833: \subsection{Time Between Primary and Secondary Transit}
834: \label{subsec:Deltat}
835:
836: In the case where the exoplanet is transiting it may be possible to
837: observe not only the primary transit, i.e.~the transit where the
838: exoplanet obscures the host star, but also the occultation, when the
839: host star blocks thermal emission and reflected light from the
840: exoplanet
841: \citep[e.g.,][]{Charb2005}.
842: %
843: If the time of the primary eclipse is given by $t_1$ and that of the
844: secondary by $t_2$, the time difference between the two as compared to
845: half a period $P$, $\Delta t \equiv t_2 - t_1 - 0.5P$, depends mostly
846: on the eccentricity $e$ and the angle of periastron $\omega$. Indeed,
847: an accurate expression that neglects terms proportional to $\cot^2 i$
848: where $i$ is the inclination angle and is therefore exact for central
849: transits, is given by
850: \be
851: \Delta t= \frac{P}{\pi}\left(\frac{e\cos(\omega)
852: \sqrt{1-e^2}}{(1-(e\sin\omega)^2)} +
853: \arctan\left( \frac{e\cos\omega}{\sqrt{1-e^2}}\right)\right)
854: \label{eq:sterne}
855: \ee
856:
857: \noindent \citep{Sterne1940}. Combined with radial velocities this time
858: difference offers an additional constrain on $e$ and $\omega$, and in
859: principle a measurement of $\Delta t$ combined with a measurement of
860: the difference in the duration of the secondary and primary eclipses
861: can be used to solve for $e$ and $\omega$ directly \citep[see
862: discussion in][]{Charb2005}. We note that \citet[][their
863: \S4.2]{Heyl2007a} also consider secondary transit timings as a means
864: to measure changes in $\omega$. While their discussion is based on
865: first order expansions in $e$ instead of using the exact expression
866: above and is phrased in different terms, it is based ultimately on the
867: same measurable quantity we discuss here.
868: \footnote{\citet{Heyl2007a} consider the time difference between
869: successive primary transits ($\Delta t_0$) and successive secondary
870: ($\Delta t_\pi$) transits, following their notation. This difference
871: can be expressed as $\Delta t_0 - \Delta t_\pi \approx -(d \Delta t /
872: d\omega) \delta \omega$, where $\Delta t$ is the quantity defined in
873: Equation~\ref{eq:sterne} and $\delta\omega$ is the change in $\omega$
874: in one orbit. }
875:
876: Equation~\ref{eq:sterne} does not include light travel time
877: contributions, i.e. it neglects the time it takes for light to travel
878: accross the system. This time is given for a central transit by
879:
880: \be
881: \Delta_{t,LT} = \frac{2a(1-e^2)}{c[1-(e\cos f_t)^2]},
882: \label{eq:LT}
883: \ee
884:
885: \noindent where $f_t$ is the true anomaly at the time of primary transit.
886: In this section we will be interested in changes in $\Delta t$ due to
887: changes in $\omega$. It is easy to see from the expressions above that
888: $d\Delta_{t,LT} / d\omega \ll d\Delta t / d\omega$ (ignoring the
889: points where they are both zero). Therefore, changes in the time
890: difference between primary and secondary transits will be dominated by
891: changes in $\Delta t$ and we can safely ignore light travel time
892: effects in what follows.
893:
894: The precession of periastra caused by GR will change $\omega$
895: systematically, while leaving $e$ unchanged. Therefore $\Delta t$ has
896: the potential of offering a direct measurement of changes in $\omega$.
897: We use Equation~\ref{eq:sterne} to calculate $(1/P) d\Delta t /
898: d\omega$ as a function of $\omega$. The result is shown in in
899: Figure~\ref{fig:dDt} for $e=0.1,0.3,0.5,0.7$, with higher $e$ yielding
900: larger extrema of $|(1/P) d\Delta t / d\omega|$.
901:
902: \begin{figure}
903: \epsscale{0.6}
904: \plotone{f4.eps}
905: \caption{
906: $(1/P) d\Delta t / d\omega$ as a function of $\omega$. The
907: different curves are for different eccentricities
908: $e=0.1,0.3,0.5,0.7$, with higher $e$ giving larger extrema of
909: $|(1/P) d\Delta t / d\omega|$. \label{fig:dDt} }
910: \end{figure}
911:
912: If $\omega \approx \pi/2$ or $\approx 3\pi/2$, the changes $\Delta t$ in
913: a system with $P=4$ days for 0.2 degrees of precession would be large,
914: on the order of 8 min for $e\sim 0.5$. Published observations of
915: secondary eclipses with {\it Spitzer} constrain $t_2$ to within
916: $\sim 80$ sec \citep{Deming2006a,Harrington2007a,Charbonneau2008a}.
917: Uncertainties in $t_2$ will dominate the uncertainty in $\Delta_t$,
918: so we assume $\sigma_{\Delta_t} \approx \sigma_{t_2} \sim
919: 80N_{tr}^{-3/2}$ sec, where $N_{tr}$ is the number of secondary
920: transits observed.
921:
922: Determing $\Delta t$ at two epochs separated by 10 years and observing
923: $N_{tr} \sim 1$ at each epoch with a facility like {\it Spitzer}
924: would detect the changes in $\Delta t$ due to GR with a significance
925: $\gtrsim 4 \sigma$ in an eccentric system that precesses $\gtrsim$ 0.2
926: deg in a decade if they have $\omega$ around $\pi/2$ or $3\pi/2$. While
927: {\it Spitzer} is now entering the end of its cryogenic lifetime, future
928: facilities might be able to achieve similar precision. Moreover, as
929: new suitable systems are discovered with very hot atmospheres
930: \citep[pM class planets,][]{Fortney2007b} observations of secondary
931: transits might be feasible from the ground
932: \citep{Lopez-Morales2007a}.
933:
934: Finally, we note that since $f_t=\pm\pi/2$ when $\omega=0$ or $\pi$,
935: measuring the change in $\Delta t$ complements measurements of
936: changes in primary transit duration $D$, in the sense that when one
937: effect is not operating the other one generally is.
938:
939: %\section{Elucidating the origin of a measured $\dot{\omega}$.}
940: \section{Assessing the Different Contributions to $\dot{\omega}$}
941: \label{sec:distinguish}
942:
943: If a change in $\omega$ is detected using any of the methods described
944: above it would be interesting to determine its likely cause. As
945: mentioned in \S~\ref{sec:prec}, the contribution to $\dot{\omega}$
946: from a second planet or tidal deformations can be of comparable
947: magnitude to that caused by GR, making the origin of a potential
948: detection of a change in $\omega$ unclear unless the presence of a
949: perturber and the magnitude of tidal effects can be assessed.
950: %
951: In this section we address three points. First, we discuss how may one
952: assess if part of a detected change in $\omega$ comes from a
953: perturber. Secondly, we show that the a terrestrial mass ($ M_\oplus
954: \lesssim M \lesssim 10$) perturber may be detectable through the
955: induced secular changes in $\omega$ over the value expected from GR
956: and tidal deformations by using transit and radial velocity
957: observations. Finally, we assess the relative contributions of GR and
958: tidal deformations for the case where perturbers do not contribute
959: significantly to $\dot{\omega}$.
960:
961: \subsection{Constraining Potential Contributions to $\dot{\omega}$
962: from $\dot{\omega}_{\rm perturber}$.}
963: \label{ssec:constr_pert}
964:
965: First, we note that using a precise radial velocity curve with $N_{\rm
966: obs} \sim 100$ we can probe the presence of companions with $M
967: \gtrsim 15 M_\oplus$ \citep{Narayan2005a}. As shown in \S~\ref{sec:prec},
968: lower mass companions can still contribute significantly to
969: $\dot{\omega}$, so it would be useful to have further information
970: about possible companions in order to better interpret a measured
971: $\dot{\omega}$.
972: %
973: The presence of a second exoplanet in the system will not only cause
974: secular changes in $\omega$ as discussed in \S~\ref{sec:prec}
975: (equation~\ref{eq:wdot_pert}), but it may cause significant
976: transit-time {\it variations} even for perturbers with masses
977: comparable to the earth \citep{Agol2005a,Holman2005a}. The
978: transit-time variations will be on the order of seconds to minutes,
979: depending on the orbital parameters and mass of the perturber.
980: Constant monitoring of transiting exoplanets such as will be performed
981: by {\it Kepler} will allow to identify systems that show significant
982: transit time variations.
983:
984: We note that the presence of a perturber will not only change $\omega$
985: but it may also affect other orbital elements such as the orbit
986: inclination
987: \citep{Miralda-Escude2002a}. These small changes in inclination have indeed
988: been proposed as a method of detecting terrestrial mass
989: companions in near-grazing transit systems which are especially
990: sensitive to them \citep{Ribas2008a}. Following the formalism
991: presented in \S6 of \citet{MurrayDermott} we expect that secular
992: changes to $e$ by a perturber satisfy $|d e / dt| \lesssim (3/16) n
993: (a/a_2)^3 (M_2/\mstar) e_2$, where $a_2, e_2$ and $M_2$ are the
994: orbital semi-major axies, eccentricity and mass of the perturber
995: respectively. For $e_2=0.1$, $a_2 = 2a$ and $M_2=3\times10^{-6}\mstar$
996: we have that $de/dt \lesssim 4 \times 10^{-6}$ yr$^{-1}$ for a system
997: with $P=4$ days. This is too small for changes in $e$ to be detected
998: with radial velocities in scales of tens of years and so we conclude
999: that secular changes in eccentricities induced by perturbers will in
1000: general not be useful to infer the presence of perturbers.
1001:
1002: Summarizing, the best way to try to constrain contributions of
1003: $\dot{\omega}_{\rm perturber}$ to a detected change in $\omega$ is to
1004: consider transiting systems posessing extensive photometric
1005: monitoring, allowing to probe the existence of transit time
1006: variations. For near-grazing systems, the same photometric monitoring
1007: would additionally allow to probe for small changes in $i$ due to a
1008: perturber.
1009:
1010:
1011: \subsection{Using Secular Variations in $\omega$ to Detect
1012: Terrestrial Mass Planets.}
1013: \label{ssec:tpl}
1014:
1015: A very interesting possibility raised by the secular variation in
1016: $\omega$ induced by a ``perturber'' is to use these variations in
1017: order to infer the presence of terrestrial mass planets. This
1018: possibility was studied by \citet{Miralda-Escude2002a} and then
1019: followed-up by \citet{Heyl2007a}. Secular variations in $\omega$,
1020: especially through their effect on transit durations
1021: (\S~\ref{subsec:D}), can offer an interesting complement to transit
1022: time variations as a means of detecting terrestrial mass planets in
1023: the upcoming {\it Kepler} mission.
1024:
1025: As we have seen in \S~\ref{sec:prec}, the precession due to GR and
1026: tidal deformations can be of comparable magnitude to that induced by a
1027: terrestrial mass perturber. It is germane to ask then to what extent
1028: will the uncertainty in the expected value of \dotw and
1029: $\dot{\omega}_{\rm tide}$ limit the detectability of a perturber.
1030: %
1031:
1032: We start by considering \dotw. The fractional uncertainty in \dotw is
1033: given by $(\sigma(
1034: \dotw) / \dotw)^2 = (\sigma( \mstar) / \mstar)^2 + (\sigma( a) / a)^2 +
1035: (\sigma( P) / P)^2 + 4e^2 (\sigma( e) / (1-e^2))^2$, where we have
1036: ignored correlations. The fractional uncertainty in $P$ is generally
1037: negligible, while the other quantities can be typically of the order
1038: of a few percent. We will therefore assume that $\sigma(\dotw) / \dotw
1039: \sim 10\%$. It follows that to detect an excess precession caused by a
1040: perturber we need at least that $\dot{\omega}_{\rm perturber} - \dotw
1041: \gtrsim 0.3\dotw$. It is easy to see that if $\dot{\omega}_{\rm
1042: perturber} \gtrsim \dotw$ and $\dot{\omega}_{\rm perturber}$ is itself
1043: detectable, i.e. $\dot{\omega}_{\rm perturber} >
1044: 3\sigma_{\dot{\omega}}$, where $\sigma_{\dot{\omega}}$ is the
1045: uncertainty in the measured $\dot{\omega}$, then the uncertainty in
1046: the expected $\dotw$ will not spoil the significance of the detection.
1047: %
1048: Using the equations presented in \S~\ref{sec:prec} we get that
1049: \be
1050: \frac{\dot{\omega}_{\rm perturber}}{\dotw} = \frac{3.8}{(1-e^2)}
1051: \left(\frac{M_\odot}{\mstar}\right)^2
1052: \left(\frac{a}{a_2}\right)^3
1053: \left(\frac{a}{0.05\mbox{AU}}\right)
1054: \left(\frac{M_2}{M_\oplus}\right).
1055: \ee
1056: It is clear from this expression that for typical values of $\mstar
1057: \sim M_\odot$, $a \sim 0.05$ AU, $e \lesssim 0.5$ there will be values
1058: of $a_2$ for which we have $\dot{\omega}_{\rm perturber} > \dotw$ and
1059: for which the uncertainty in the expected precession from GR will not
1060: be a limiting factor in detecting terrestrial mass perturbers.
1061:
1062: We consider now $\dot{\omega}_{\rm tide}$. The fractional uncertainty
1063: considering all parameters excepting $k_{2,p}$ and ${\cal T}$ and ignoring
1064: correlations is $(\sigma(\dot{\omega}_{\rm tide}) / \dot{\omega}_{\rm
1065: tide})^2 = (\sigma( P) / P)^2 + (5\sigma( \rpl/a) / (\rpl/a))^2 +
1066: (\sigma(
1067: \mpl/\mstar) / (\mpl/\mstar))^2 + (f^\prime(e)
1068: \sigma(e)/f(e))^2$. We have expressed this in terms of the ratios
1069: $\mpl/\mstar$ and $\rpl/a$ as these quantities are more robustly
1070: determined observationally. As was the case above, the uncertainties
1071: in the variables can be typically a few percent and we therefore
1072: assume that $\sigma(\dot{\omega}_{\rm tide}) /
1073: \dot{\omega}_{\rm tide} \sim 10\%$.
1074: Using the equations presented in \S~\ref{sec:prec} we get that
1075: \be
1076: \frac{\dot{\omega}_{\rm perturber}}{\dot{\omega}_{\rm tide}} =
1077: \frac{1.85}{{\cal T} f(e)k_{2,p}}
1078: \left(\frac{M_\odot}{\mstar}\right)^2
1079: \left(\frac{a}{a_2}\right)^3
1080: \left(\frac{a}{0.05\mbox{AU}}\right)^5
1081: \left(\frac{R_J}{\rpl}\right)^5
1082: \left(\frac{\mpl}{M_J}\right)
1083: \left(\frac{M_2}{M_\oplus}\right).
1084: \ee
1085: In the case of $\dot{\omega}_{\rm tide}$ we cannot measure the apsidal
1086: motion constant $k_2$. As discussed in \S~\ref{sec:prec}, the value of
1087: $k_2$ for a giant planet is expected to be close to the extreme value
1088: of a uniform sphere, so we can conservatively assume $k_{2,p}=0.75$, a
1089: value that maximizes the expected $\dot{\omega}_{\rm tide}$. For a
1090: close-in Jupiter we have ${\cal T} \approx 1$. Using these values and
1091: following the same reasoning as for \dotw, it is clear from the
1092: expression above that for typical values of $\mstar
1093: \sim M_\odot$, $a \sim 0.05$ AU, $e \lesssim 0.5$, $\rpl \sim R_J$,
1094: $\mpl \sim M_J$, there will be values of $a_2$ for which we have
1095: $\dot{\omega}_{\rm perturber} >\dot{\omega}_{\rm tide} $ and for which
1096: the uncertainty in the expected precession arising from tidal
1097: deformations will not be a limiting factor in detecting terrestrial
1098: mass perturbers. Note that due to the strong dependence on $a$, tides
1099: may become a limiting factor for very close-in systems.
1100:
1101: We consider now the detectability of $\dot{\omega}_{\rm perturber}$
1102: with {\it Kepler} using the change in the transit duration $D$. As
1103: discussed in \S~\ref{subsec:D}, a 1-minute sampling of a Jupiter-mass
1104: system with {\it Kepler} for a $P=5$ days planet orbiting a $V=12$
1105: solar-type star will achieve a precision of $\sigma_D \sim 1.5$ sec
1106: for a 0.15 days transit duration. We therefore set a change of 6 sec
1107: over 4 years to constitute a detectable $\delta D$ for this system,
1108: which translates into a detectable $\dot{\omega}$ of
1109: $\dot{\omega}_{\rm detect} = 6 / (D
1110: \Delta t (\dot{\omega}^{-1}d\ln D/dt))$, where $\Delta t = 4$ years and
1111: we assume a central transit (see Equation~\ref{eq:dlnD}).
1112:
1113: We show in Figure~\ref{fig:plan_kep} the expected value of
1114: $\dot{\omega}_{\rm perturber}$ as a function of $a_2/a$ assuming
1115: $a=0.05$ AU, $\mstar=M_\odot$, and $f_t=0.5\pi$ as solid lines, one
1116: for $M_2=M_\oplus$ and another for $M_2=10M_\oplus$. The dashed line
1117: marks the values of $\dot{\omega}_{\rm detect}$ for $e=0.4$. The
1118: dotted lines mark the values of $0.3\dotw$ for the same eccentricity,
1119: while the dash-dot-dot line marks $0.3\dot{\omega}_{\rm tide}$ This
1120: figure shows that over a range of values of $(a_2/a)$ {\it Kepler} may
1121: be able to detect the presence of additional super-Earths ($M_\oplus
1122: \lesssim M \lesssim 10M_\oplus$) if the primary planet has a favorable
1123: impact parameter and true anomaly at the time of transit.
1124: %
1125:
1126: Finally, we consider the detectability of $\dot{\omega}_{\rm
1127: perturber}$ using radial velocities alone. We assume the primary
1128: planet is a super-massive system of $\mpl = 10 M_J$, with the rest of
1129: the star and orbital parameters as assumed above. The results of
1130: \S~\ref{subsec:rv} show that with 100 radial velocity observations at
1131: each epoch a precession of $\sim 8$ deg/century could be detected (at
1132: $3\sigma$) in 5 years if the system has $e=0.4$ and orbits an inactive
1133: star. We mark the detectable level in this case with radial velocities
1134: as dot-dashed line in Figure~\ref{fig:plan_kep}. From this Figure we
1135: can see that if enough observations are in hand, perturbing planets
1136: with masses $\sim 10 M_\oplus$ may be detectable with radial
1137: velocities alone for the case of super-massive systems orbiting
1138: inactive stars.
1139:
1140: \begin{figure}
1141: \epsscale{0.6}
1142: \plotone{f5.eps}
1143: \caption{
1144: The solid lines show the expected value of $\dot{\omega}_{\rm
1145: perturber}$ as a function of $a_2/a$ for $M_2=M_\oplus$ and
1146: $M_2=10M_\oplus$, assuming $a=0.05$ AU, $\mstar=M_\odot$ and
1147: $f_t=0.5\pi$. The dashed line marks the lower value of
1148: $\dot{\omega}$ that is detectable with {\it Kepler} for
1149: $e=0.4$ via the changes in transit duration. The dotted line
1150: marks the value of $0.3\dotw$ for the same eccentricity, while
1151: the dash-dot-dot line marks $0.3 \dot{\omega}_{\rm
1152: tide}$. Finally, the dot-dashed line marks the lower values of
1153: $\dot{\omega}$ that would be detectable with 100 radial
1154: velocities at each of two epochs separated by 5 years, for a
1155: perturber to a super-massive $M_1=10M_J$ primary planet
1156: orbiting an inactive solar-mass star (see text for more
1157: details). \label{fig:plan_kep} }
1158: \end{figure}
1159:
1160: We conclude that measuring the precession of $\omega$ caused by a
1161: perturber above the value predicted by GR and tidal effects may lead
1162: to detection of additional planets with both transit and radial
1163: velocity observations (for transits, see also
1164: \citet{Miralda-Escude2002a,Heyl2007a}). The examples given above are
1165: illustrative only; a detailed analysis is beyond the scope of this
1166: work. Of special interest will be to estimate the yield of
1167: terrestrial-mass planets expected from {\it Kepler} by measuring
1168: secular changes in transit durations.
1169:
1170: \subsection{The Relative Magnitude of \dotw and $\dot{\omega}_{\rm tide}$.}
1171: \label{ssec:rel_mag}
1172:
1173: If there are no perturbers causing significant changes in $\omega$,
1174: these changes will be caused by a combination of GR and tidal
1175: deformations. We now briefly consider the relative magnitudes of these
1176: effects. Using the equations presented in \S~\ref{sec:prec} we find
1177: that
1178: \be
1179: \frac{\dotw}{\dot{\omega}_{\rm tide}} =
1180: \frac{4.8}{{\cal T} (1-e^2) f(e)}
1181: \left(\frac{0.1}{k_{2,p}}\right)
1182: \left(\frac{a}{0.05\mbox{AU}}\right)^4
1183: \left(\frac{R_J}{\rpl}\right)^5
1184: \left(\frac{\mpl}{M_J}\right).
1185: \ee
1186: For expected values of $k_{2,p} \sim 0.25$ and assuming $\rpl \sim
1187: R_J$, $\mpl \sim M_J$, $a\sim 0.05$ and $e\lesssim 0.5$ we see that
1188: the values of \dotw and $\dot{\omega}_{\rm tide}$ are comparable. Note
1189: that there is a very strong dependence on $a$. For smaller values of
1190: $a$, a regime where the precession due to GR would offer better
1191: chances of being observable for eccentric systems, the contribution
1192: from tidal deformations may quickly become the dominant source of
1193: precession. The fact that the tidal contributions to the precession
1194: are expected to be generally significant limits the ability to
1195: directly extract the precession due to GR.
1196: %
1197: The magnitude of the precession caused by tidal deformations is
1198: uncertain due to the need to know the planetary structure. Given that
1199: the expected precession due to GR is unambigous, in systems where the
1200: precession due to tidal deformations is expected to be dominant {\it
1201: and} detectable, we might be able to learn about the tidally driven
1202: precession of $\omega$ by subtracting the effects of GR from a
1203: measured $\dot{\omega}$.
1204:
1205: \section{Analysis of Specific Super-Massive Systems}
1206:
1207: As illustration of the material discussed above, we analyze now in
1208: some detail two currently known systems that due to their large
1209: measured velocity amplitudes have the potential to have a measurable
1210: $\dot{\omega}$ using all the techniques presented above: HAT-P-2 b
1211: \citep{Bakos2007a,Loeillet2008a} and XO-3 b
1212: \citep{Johns-Krull2008a}.
1213:
1214: \subsection{HAT-P-2 b}
1215:
1216: HAT-P-2 b is an especially interesting system due to its very high
1217: eccentricity, high mass of $\approx 9 M_{J}$ and close orbit
1218: ($a=0.0677$ AU) to its host star. As shown in Table~\ref{tab:prec} its
1219: value of \dotw is $\sim 2^\circ$/century, while the expected value of
1220: $\dot{\omega}_{\rm tide}$ is $\sim 0.25$ deg/century assuming
1221: $k_{2,p}=0.25$. Unfortunately it has a high level of stellar jitter,
1222: with estimates ranging from 17 m sec$^{-1}$ to 60 m sec$^{-1}$
1223: \citep{Bakos2007a,Loeillet2008a}. So even though this system has
1224: $K\sim$ 1000 m sec$^{-1}$ it would require an unrealistically large
1225: number of observations ($\sim 10^4$) per epoch in order to make its
1226: expected \dotw detectable with radial velocities.
1227:
1228: The value of $\omega$ derived for HAT-P-2 b is $1.05\pi$ and we
1229: therefore would not expect to see significant variations in the time
1230: between primary and secondary transits in case the secondary transit
1231: was observed (see Figure~\ref{fig:dDt}).
1232: %
1233: With $\omega = 1.05\pi$ the true anomaly at the time of transit will
1234: be $f_t \approx 0.45 \pi$, or $1.55\pi$. Given that the impact
1235: parameter is $b\approx 0$ we get from Equation~\ref{eq:dlnD} that
1236: $|d\ln D/dt| \sim 1.67\times10^{-2}$ century$^{-1}$. Using the fact
1237: that $D= 0.15$ days we expect then a change of $\sim 21$ sec in $D$
1238: over 10 years due to GR. This difference could be readily detected
1239: with high precision photometric observations that determine $D$ to
1240: within a few seconds such as is possible with {\it HST}. The
1241: observations currently available constrain $D$ only to within $\sim 3$
1242: mins \citep{Bakos2007a}, and so no first epoch suitable to measuring
1243: changes in $D$ is yet in hand.
1244:
1245: \subsection{XO-3 b}
1246: \label{ssec:XO3}
1247:
1248: As shown in Table~\ref{tab:prec} XO-3 b has the largest predicted
1249: \dotw of all currently known exoplanets with $e>0.1$. While its
1250: eccentricity is not as high as that of HAT-P-2 b, it is more massive
1251: and orbits closer to its host star XO-3 (also known as
1252: GSC~03727-01064). Asuming $k_{2,p}=0.25$, the expected value of
1253: $\dot{\omega}_{\rm tide}$ is $\sim 11$ deg/century, about three times
1254: as much as the contribution from GR.
1255:
1256: In order to assess the detectability of $\dot{\omega}$ with radial
1257: velocities for XO-3 b we need to know $\sigma_{\rm jitter}$, but
1258: unfortunately the precision of the radial velocity measurements
1259: presented in \citet{Johns-Krull2008a} is too coarse to allow a
1260: determination of this quantity ($\sigma_{\rm obs} \gtrsim 100$ m
1261: sec$^{-1}$). Based on the spectral type F5V and $v\sin i=18.5\pm 0.2$
1262: km sec$^{-1}$ of XO-3 b \citep{Johns-Krull2008a} we can expect it to
1263: have a rather high value of stellar jitter $\sigma_{\rm jitter}
1264: \gtrsim 30$ m sec$^{-1}$ \citep{Saar1998a}. Therefore, and just as is
1265: the case for HAT-P-2 b, we do not expect $\dot{\omega}$ to be
1266: detectable with radial velocity observations due to the expected
1267: jitter.
1268:
1269: The value of $\omega$ derived for XO-3 b is consistent with 0 and we
1270: therefore would not expect to see variations in the time between
1271: primary and secondary transits in case the secondary transit was
1272: observed (see Figure~\ref{fig:dDt}). As $\omega \approx 0$, the true
1273: anomaly at the time of transit will be $f_t \approx \pi/2$ or
1274: $3\pi/2$. Given that the impact parameter is $b\approx 0.8$ we get
1275: from Equation~\ref{eq:dlnD} that $|d\ln D/dt|
1276: \sim 3.56\times10^{-2}$ century$^{-1}$. Using the fact that
1277: $D = 0.14$ days we expect then a change of $\sim 43$ sec in $D$ over
1278: 10 years. Just as is the case for HAT-P-2 b, this could be
1279: detectable with determinations of $D$ to within a few seconds and
1280: there is no suitable first epoch yet in hand.
1281:
1282: \section{Conclusions}
1283:
1284: In this work we have studied the observability of the precession of
1285: periastra caused by general relativity in exoplanets. We additionally
1286: consider the precession caused by tidal deformations and planetary
1287: perturbers, which can produce a precession of comparable or greater magnitude.
1288: %
1289: We consider radial velocities and transit light curve observations and
1290: conclude that for some methods precessions of the magnitude expected
1291: from GR will be detectable in timescales of $\sim 10$ years or less
1292: for some close-in, eccentric systems. In more detail, we find that:
1293:
1294: \begin{enumerate}
1295:
1296: \item For transiting systems, precession of periastra of the magnitude
1297: expected from GR will manifest itself through detectable changes in
1298: the duration of primary transit (\S\ref{subsec:D}) or through the
1299: change in the time between primary and secondary transits
1300: (\S\ref{subsec:Deltat}) in timescales of $\lesssim 10$ years. The two
1301: methods are most effective at different values of the true anomaly. A
1302: determination of the primary transit duration time to $\sim$ a few
1303: seconds and that of the secondary to $\sim$ a minute will lead to
1304: measurable effects. The effects of GR and tidal deformations might
1305: need to be included in the analysis of {\it Kepler} data for
1306: eccentric, close-in systems. The transit duration of near-grazing
1307: systems will be particulary sensitive to changes in $\omega$.
1308:
1309:
1310: \item Radial velocity observations alone would be able to detect
1311: changes in the longitude of periastron of the magnitude expected from
1312: GR effects only for eccentric super-massive ($K \sim 1000$ m
1313: sec$^{-1}$) exoplanets orbiting close to a host star with a low-level
1314: of stellar jitter (\S\ref{subsec:rv}). For the detection to be
1315: statistically significant, on the order of $100$ precise radial
1316: velocity observations are needed at each of two epochs separated by
1317: $\sim 20$ years.
1318:
1319: \item Measurements of the change over time of the period between
1320: primary transits is not currently a method that will lead to a
1321: detection of changes in $\omega$ of the magnitude expected from GR
1322: (\S\ref{subsec:dPdt}). Previous works have shown that measuring the
1323: small difference between the radial velocity period and that of
1324: transits are not sensitive enough to lead to detectable changes due to
1325: GR.
1326:
1327: \end{enumerate}
1328:
1329:
1330: In order to contrast any detected change in the transit duration
1331: (\S\ref{subsec:D}) or the time between primary and secondary
1332: (\S\ref{subsec:Deltat}) to the predictions of a given mechanism one
1333: needs to know the eccentricity and longitude of periastron of the
1334: systems, for which radial velocities are needed (although not
1335: necessarily of the precision required to directly detect changes in
1336: $\omega$ with them\footnote{The eccentricity can be constrained using
1337: transit information alone, see \citet{Ford2008a}}).
1338: %
1339: Conversely, photometric monitoring of primary transits are useful in
1340: order to elucidate the nature of a detected change in $\omega$ by
1341: probing for the presence of transit time variations. The presence of
1342: the latter would imply that at least part of any observed changes in
1343: $\omega$ could have been produced by additional planetary companions
1344: (\S\ref{sec:distinguish}).
1345:
1346: Precession of periastra caused by planetary perturbers and the effects
1347: of tidal deformations can be of comparable magnitude to that caused by
1348: GR (\S\ref{ssec:add_prec}). The effects of tidal deformations on the
1349: precession of periastra in particular may be of the same magnitude or
1350: dominate the total $\dot{\omega}$ in the regime where the GR effects
1351: are detectable (\S\ref{ssec:rel_mag}). While this limits the ability
1352: to directly extract the precession due to GR given the uncertainty in
1353: the expected precession from tides, it might allow to study the
1354: tidally induced precession by considering the residual precession
1355: after subtracting the effects of GR. The latter possibility is
1356: particularly attractive in systems where the tidally induced
1357: precession may dominate the signal.
1358: %
1359: We note that even without considering the confusing effects of tidal
1360: contributions to the precession, a measurement of \dotw as described
1361: in this work would not be competitive in terms of precision with
1362: binary pulsar studies \citep[see, e.g.,][for a review ]{Will2006a} and
1363: would therefore not offer new tests of GR.
1364:
1365: The upcoming {\it Kepler} mission expects to find a large number of
1366: massive planets transiting close to their host stars
1367: \citep{Borucki2003a}, some of which will certainly have significant
1368: eccentricities. Furthermore, systems observed by {\it Kepler} will be
1369: extensively monitored for variations in their transiting time periods
1370: in order to search for terrestrial-mass planets using transit-time
1371: variations. We have shown that modeling of the transit time durations
1372: and further characterization of close-in, eccentric systems might need
1373: to take into account the effects of GR and tidal deformations as they
1374: will become detectable on timescales comparable to the 4-year lifetime
1375: of the mission, and certainly on follow-up studies after the mission
1376: ends. We have also shown that planetary companions with super-Earth
1377: masses may be detectable by {\it Kepler} by the change in transit
1378: durations they induce (\S~\ref{ssec:tpl}). Additionally, well sampled
1379: radial velocity curves spanning $\gtrsim 5$ years may also be able to
1380: detect companions with super-earth massses by measuring a change in
1381: $\omega$ over the expected GR value for the case of super-massive,
1382: close-in systems orbiting inactive stars (\S~\ref{ssec:tpl}).
1383:
1384:
1385: \acknowledgements
1386:
1387: We would like to thank the anonymous referee for helpful suggestions
1388: and Dan Fabrycky and Andr\'as P\'al for useful
1389: discussions. G.B. acknowledges support provided by the National
1390: Science Foundation through grant AST-0702843.
1391:
1392: \bibliographystyle{apj}
1393: \bibliography{allrefs}
1394:
1395:
1396:
1397:
1398:
1399:
1400: \end{document}
1401: