HD 285507b: AN ECCENTRIC HOT JUPITER IN THE HYADES OPEN CLUSTER

The efforts of more than two decades of exoplanet searches have produced an incredible diversity of discoveries, many of which bear little similarity to planets in the solar system. These discoveries provide valuable constraints for our theories of planet formation, which have evolved to explain the existence of the wide array of planetary systems observed. One such grouping of planets is the hot Jupiters, gas giant planets in short period orbits (defined herein as M_P > 0.3 M_Jup and P < 10 days), which occur at rates of ∼1.2% around Sun-like field stars (see Wright et al. 2012; Mayor et al. 2011). However, more than 60 years after their proposed existence (Struve 1952) and nearly 20 years after the first detection (Mayor & Queloz 1995), we still lack a complete description of the hot Jupiter formation process. It is believed that they form beyond the snow line (located at ∼2.7 AU for the current-day Sun; e.g., Martin & Livio 2013) where there is a greater reservoir of solids with which to build a massive core (e.g., Kennedy & Kenyon 2008) before undergoing an inward migration process, but many questions remain to be answered.

While there are many mechanisms that could cause a gas giant planet to lose angular momentum and migrate inward (see a discussion in Ford & Rasio 2008), two leading ideas are dynamical interactions with a circumstellar disk ("Type II" migration; Goldreich & Tremaine 1980; Lin & Papaloizou 1986) and multi-body gravitational interactions with other planetary ("planet–planet scattering"; e.g., Rasio & Ford 1996; Juric & Tremaine 2008) or stellar (Kozai cycles; e.g., Fabrycky & Tremaine 2007) companions. Migration through a disk must occur while the gas disk is present (within ∼10 Myr; Haisch et al. 2001) and is expected to preserve near-circular orbits, but if multi-body interactions are the source of most hot Jupiters, inward migration may take significantly longer and many of these planets should initially possess high eccentricity. For simplicity, we adopt the language of Socrates et al. (2012) and refer to these multi-body processes as "high eccentricity migration" (HEM). Given the different timescales predicted, one direct way to distinguish between mechanisms would be to search for hot Jupiters orbiting very young stars (≲ 10 Myr), and at least one promising candidate exists orbiting a T-Tauri star (PTFO 8-8695; van Eyken et al. 2012; Barnes et al. 2013). However, the enhanced activity associated with young stars presents significant challenges for such surveys (e.g., Bailey et al. 2012). Alternatively, the dynamical imprint of HEM on the orbital eccentricity could provide a more accessible means to observationally constrain migration (e.g., Dawson & Murray-Clay 2013; Dong et al. 2014). Subsequent tidal circularization of the orbits erases this evidence of multi-body interaction over time, so identifying "dynamically young" systems, for which the system age (t_age) is less than the circularization timescale (τ_cir), is necessary for such an investigation. Longer period planets (large τ_cir) could satisfy this requirement, but hot Jupiters with periods of only a few days are both more common and easier to detect. Young planets (small t_age) offer another solution, but field stars tend to be old and their ages are difficult to estimate accurately (Mamajek & Hillenbrand 2008). The ages of open clusters, on the other hand—e.g., the 625 Myr Hyades (Perryman et al. 1998) or the 125 Myr Pleiades (Stauffer et al. 1998)—are typically precisely known and can be comparable to (or less than) the circularization timescales of many hot Jupiters. Consequently, planets in clusters could provide an avenue to directly measure this observational signature of HEM, allowing us to determine which process is most important for hot Jupiter migration.

With the recent discovery of two hot Jupiters (Quinn et al. 2012) in the Praesepe open cluster (∼600 Myr), it is now clear that giant planets can form and migrate in a dense cluster environment. This had been an open question, as previous exoplanet surveys targeting open clusters—including high resolution spectroscopy of 94 dwarfs in the Hyades (Paulson et al. 2004, hereafter P04) and 58 dwarfs in M67 (Pasquini et al. 2012), as well as transit photometry of several other clusters (e.g., Hartman et al. 2009; Pepper et al. 2008; Mochejska et al. 2006)—had not discovered any planets. However, while small planets do appear to be a common by-product of star formation, giant planets are comparatively rare (e.g., Fressin et al. 2013). Most cluster surveys have not been sensitive to the more common smaller planets, so it is likely that small sample sizes are to blame for the null results (see also van Saders & Gaudi 2011). Indeed, two mini-Neptunes recently discovered (Meibom et al. 2013) in NGC 6811 (∼1 Gyr, Meibom et al. 2011) by the Kepler spacecraft would not have been detected by previous surveys, and the results suggest consistency between occurrence rates in open clusters and the field for planets of all sizes.

While open cluster planet surveys have recently enjoyed successes in measuring planet formation rates, they have not yet provided a convincing constraint for the hot Jupiter migration process, despite their potential to do so in the long term. Unfortunately, the hot Jupiters discovered in Praesepe are too old to address timescale differences between migration mechanisms, and they are not dynamically young, either. That is, they are close enough to their stars to have already undergone tidal circularization (t_age > τ_cir). To see the dynamical imprint of migration it will thus be necessary to identify younger hot Jupiters or hot Jupiters on wider orbits that would still bear the signature of dynamical scattering in order to help distinguish between migration mechanisms. In hopes of accomplishing the latter, here we extend our radial velocity (RV) survey to include more stars with ages and properties comparable to those in Praesepe.

We describe our sample selection in Section 2 and outline our spectroscopic and photometric observations and analysis in Sections 3 and 4. We then present evidence for high eccentricity migration and a constraint on Q_P in Sections 5–7, and in Section 8 we discuss our results.

Hyades cluster members represent an ideal sample for extending our survey of Sun-like stars in Praesepe (Quinn et al. 2012). The two clusters are nearly the same age (625 Myr and 578 Myr, respectively; Perryman et al. 1998; Delorme et al. 2011), and both are metal-rich—[Fe/H] = +0.13 ± 0.01 (Hyades, Paulson et al. 2003) and +0.19 ± 0.04 (Praesepe, Quinn et al. 2012). While P04 found no planets among 94 Hyades dwarfs (74 FGK stars), a primary motivation of their survey was to explore the relationship between planet occurrence and stellar mass, so a number of FGK stars were omitted in favor of including 20 M stars. As a result, some suitable stars have not yet been observed with precise radial velocities.

Our potential targets come from a list assembled by Robert Stefanik for the CfA Hyades binary survey (R. Stefanik 2013, private communication). Since 1979, the CfA has been monitoring over 600 stars in the Hyades field, extending to a magnitude of V = 15. Originally the program consisted of stars drawn from the Hyades lists of van Bueren (1952), van Altena (1966), and Pels et al. (1975). Over the years additional stars were added to the observing program if there was some suggestion that the stars were possible members based on photometric, proper motion, or radial velocity investigations of the cluster. Also added to the CfA program were the companion stars of Hyades visual binaries. From this parent list, the sample surveyed here was determined after excluding close (<10) visual binaries revealed by high-resolution imaging (Patience et al. 1998), the 94 stars previously surveyed by P04, rapid rotators (vsin (i) > 30 km s⁻¹), faint targets (V > 12), and spectroscopic binaries (with orbits or long-term trends) from the literature and the CfA survey. The final target list contained 27 FGK Hyades members (see Table 1).

3.1. Spectroscopic Observations

3.2. Spectroscopic Reduction and Cross-correlation

3.3. Orbital Solution

We used a Markov Chain Monte Carlo (MCMC) analysis to fit Keplerian orbits to the radial velocity data of HD 285507, fitting for orbital period P, time of conjunction T_c, the radial velocity semi-amplitude K, the center-of-mass velocity γ_rel, and the orthogonal quantities $\sqrt{e}\cos {\omega }$ and $\sqrt{e}\sin {\omega }$, where e is eccentricity and ω is the argument of periastron. We adopted errors corresponding to the extent of the central 68.3% interval of the MCMC posterior distributions.

The RV errors did not require the addition of stellar jitter in order to obtain a good fit ($\chi ^2_{\rm red}=1$), so we set σ_* = 0 in the orbital solution. We report the best fit orbital parameters in Table 3 and plot the best fit orbit in Figure 1.

Zoom In Zoom Out Reset image size

Table 3. Stellar and Planetary Properties

Orbital parameters
P (days)	6.0881 ± 0.0018
Tc (BJD)	2456263.121 ± 0.029
K (m s−1)	125.8 ± 2.3
ea	0.086 ± 0.019
ω (deg)a	182 ± 11
γrel (m s−1)	143.9 ± 1.6
γabs (km s−1)b	38.149 ± 0.023
Physical properties
M* (M☉)c	0.734 ± 0.034
R* (R☉)c	0.656 ± 0.054
Teff, * (K)c	$4503^{+85}_{-61}$
log g* (dex)c	$4.670^{+0.051}_{-0.058}$
vsin i (km s−1)	3.2 ± 0.5
[Fe/H] (dex)c	+0.13 ± 0.01
Age (Myr)c	625 ± 50
MPsin i (MJup)	0.917 ± 0.033

Notes. ^aThe MCMC jump parameters included the orthogonal quantities $\sqrt{e} \cos {\omega }$ and $\sqrt{e} \sin {\omega }$, but we report the more familiar orbital elements e and ω. ^bThe absolute center-of-mass velocity has been shifted to the RV scale of Nidever et al. (2002), on which the velocities of HD 3765 and HD 38230 are −63.202 km s⁻¹ and −29.177 km s⁻¹, respectively. ^cFrom the final isochrone fits (Section 3.5). [Fe/H] and age were fixed to values determined for the cluster (Paulson et al. 2003; Perryman et al. 1998).

Download table as:  ASCIITypeset image

Because a modest non-zero eccentricity causes only a small deviation from a circular orbit, we also investigated whether there exists correlated RV noise (e.g., due to surface activity and rotation) on timescales similar to the orbital period using the method of Winn et al. (2010). Such noise could in principle cause small deviations from a circular orbit that might be interpreted as orbital eccentricity. To rule out this scenario, we fit a circular orbit and performed the test on the residuals to that solution. We found no evidence for correlated noise on any timescale.

3.4. Tests for a False Positive

HD 285507 is slowly rotating (P_rot = 11.98 days; Delorme et al. 2011), no X-ray emission was detected by ROSAT (Stern et al. 1995), and no stellar jitter term was required to obtain a good fit to the radial velocities, all of which are suggestive of a chromospherically inactive star. Nevertheless, to rule out false positive scenarios in which the observed RV variations are caused by stellar activity or stellar companions, we used our observations of HD 285507 to search for spectroscopic signatures that correlate with the orbital period.

If the RV variations were caused by a background blend (Mandushev et al. 2005) or star spots (Queloz et al. 2001), we would expect the shape of the star's line bisector to vary in phase with the radial velocities. A standard prescription for characterizing the shape of a line bisector is to measure the relative velocity at its top and bottom; this difference is referred to as a line bisector span (see, e.g., Torres et al. 2005). To test against background blends or star spots, we computed the line bisector spans for all observations of HD 285507. As illustrated in Figure 1, the bisector span variations are small (σ_BS = 15 m s⁻¹) and they are not correlated with the observed RV variations, having a Pearson r value of only 0.06.

For each spectrum we also computed the S index—an indicator of chromospheric activity in the Ca ii H&K lines. We follow the procedure of Vaughan et al. (1978), but we note that our S indices are not calibrated to their scale; these are relative measurements. Correlation between S index and orbital phase might be expected if the apparent RV variations were activity-induced, but as shown in Figure 1, there is no such correlation (Pearson r = 0.17). Instead, there may be significant periodicity in the S indices at 12 or 13 days (Figure 2), which is similar to the published rotation period of 11.98 days. Our data set is too sparse to claim a detection of the rotation period from the activity measurement, but we can see there is no power at the observed orbital period of 6.088 days.

Zoom In Zoom Out Reset image size

Finally, if spots were the source of the variation, we might also expect the RV amplitude to be wavelength dependent because contrast between the spot and the stellar photosphere is wavelength dependent. We derived RVs for the blue and red orders separately (weighted mean wavelengths of λ_blue = 4967 Å and λ_red = 5845 Å), and find the amplitudes to be consistent at the level of 0.5σ (3 m s⁻¹, Figure 1). This agreement between the red and blue amplitudes is encouraging, but is not conclusive by itself. We generally expect large amplitude differences between the optical and infrared for spot-induced RVs because the spot contrast can change drastically over that wavelength range. The expected amplitude difference for our smaller (∼1000 Å) wavelength span, on the other hand, is more uncertain because the local wavelength dependence of the RV amplitudes is itself dependent upon the (unknown) temperature difference between spot and photosphere (e.g., Reiners et al. 2010; Barnes et al. 2011). The simulations of Reiners et al. (2010) indicate that the amplitude difference might be detectable if the spot contrast is low, but not if it is high. Even this is not certain, though, as other authors (Desort et al. 2007) predict a 10% drop in amplitude between blue and red for high contrast spots on solar-type stars. Regardless, from these results and the work of Saar & Donahue (1997), we estimate that to induce the observed RV amplitude (125 m s⁻¹) given vsin i ≈ 3.2 km s⁻¹, the spot would have to cover ∼20% of the visible stellar surface for low contrast spots ($\Delta _{T_{{\rm eff}}} \approx 200$ K) or ∼5% for high contrast spots ($\Delta _{T_{{\rm eff}}} \approx 1500$ K). Large and high contrast spots are more likely to appear on very magnetically active stars (e.g., Bouvier et al. 1995), and we have no evidence for strong magnetic activity. Furthermore, for such spot configurations, the RV-bisector correlation should be strong (e.g., Mahmud et al. 2011), and we observe no correlation.

We conclude from the evidence presented above that the observed RV variation is not caused by spots, but is the result of an orbiting planetary companion.

3.5. Stellar and Planetary Properties

Since the rotation period of HD 285507 is 11.98 days and we have estimates for R_* and vsin i, we can in principle calculate the inclination of the stellar spin axis. In practice, the fractional uncertainty on vsin i is large (not because the absolute uncertainty is large, but because the value is small), and the inclination can only be constrained to be i > 72°. This does not exclude an edge-on stellar equator (i = 90°), and because hot Jupiters orbiting cool stars (≲ 6250 K) tend to be well-aligned with the stellar spin axis (see, e.g., Albrecht et al. 2012), an inclination of ∼90° would make a transit more likely a priori. However, even an inclined stellar spin axis would not preclude transits of HD 285507, as there is evidence to suggest that young planets tend to be more misaligned than old planets (e.g., Triaud 2011). In addition to providing a radius measurement for HD 285507b, transits of this relatively young planet orbiting a cool star could be valuable to the interpretation of these intriguing correlations.

With this in mind, we conducted photometric monitoring of HD 285507 with KeplerCam on the FLWO 1.2 m telescope at the predicted time of conjunction on UT 2012 November 7. KeplerCam is a monolithic, 4k×4k Fairchild 486 chip with a 23' × 23' FOV and a resolution of 0336 pixel⁻¹. We used a Sloan i filter with exposure times of 45 s and readout time of 12 s, obtaining a total of 334 images over 5.5 hr. We reduced the raw images using the IRAF package mscred, and performed aperture photometry with SExtractor (Bertin & Arnouts 1996).

The resulting light curve showed no sign of a transit, but to determine our detection sensitivity, we simulated transits—using the routines of Mandel & Agol (2002) and a quadratic limb darkening law from Claret et al. (2012)—and injected them into our observed data. For each injected transit, we compared the mean flux in transit (for all points between mid-ingress and mid-egress) to the mean flux out of transit. If the two differed by more than 1σ, we classified that transit as detected. From this, we can rule out a central transit of objects larger than 0.35 R_Jup (see Figure 3). If HD 285507b were to transit, then the derived minimum mass would be the true planetary mass. Under this assumption and using the mass-radius-flux relation from Weiss et al. (2013), we would then expect its radius to be ∼0.95 R_Jup, much larger than our sensitivity limit. We can also rule out all transits of a 0.95 R_Jup planet with impact parameters b < 1 (i.e., all but the most extreme grazing transits). Using the final ephemeris, the transit center should have occurred 2.22 hr after our observations began (T_c = 2456238.7686 BJD), with an uncertainty of ∼1 hr, so it is unlikely that a transit occurred outside our observing window.

Zoom In Zoom Out Reset image size

The discovery of a hot Jupiter in the Hyades open cluster brings the total number of short period giant planets in clusters to 3. Of these, however, HD 285507b is unique in that it is the only one that is definitively eccentric; the two planets in Praesepe are consistent with having circular orbits. As described in the introduction, the non-zero eccentricity of HD 285507b could be a tracer of its migration history if the planet is dynamically young (i.e., τ_cir > t_age = 625 Myr). If it is dynamically old, then the orbit should have already circularized, and establishing a credible link between the eccentricity and the migration process becomes more difficult. In planet–planet scattering for example, if the outer planet gets ejected during the scattering event as is expected, then one must invoke a separate mechanism to excite eccentricity again after circularization. To put it differently, if HD 285507b is dynamically young, then planet-planet scattering is sufficient (but not necessary) to explain the observations; if the planet is dynamically old, planet–planet scattering is neither sufficient nor necessary. To test these scenarios, we estimate τ_cir using the equation given by Adams & Laughlin (2006):

$$\begin{eqnarray} \tau _{\rm cir} = & 1.6 \,{\rm Gyr} \times \left(\frac{Q_{\rm P}}{10^6}\right) \times \left(\frac{M_{\rm P}}{M_{\rm Jup}}\right) \times \left(\frac{M_*}{{M_{\odot }}}\right)^{-1.5} \nonumber \\ & \times \left(\frac{R_{\rm P}}{R_{\rm Jup}}\right)^{-5} \times \left(\frac{a}{0.05 \,{\rm AU}}\right)^{6.5} \end{eqnarray} \tag{ 1 }$$

where Q_P is the planetary tidal quality factor (a measure of the efficiency of tidal dissipation within the planet). Note that τ_cir scales linearly with Q_P, which is unknown to within an order of magnitude. The Jupiter–Io interaction does provide the constraint 6 × 10⁴ < Q_Jup < 2 × 10⁶ (Yoder & Peale 1981), but Q_P is likely dependent upon temperature, composition, rotation, and internal structure, all of which may be quite different for hot Jupiters. Q_P ≈ 10⁶, which we adopt herein, is a fiducial value often assumed for short period giant planets (for a more detailed discussion of tidal dissipation, see, e.g., Ogilvie & Lin 2004).

Since HD 285507b does not transit, we do not know R_P and measure only a minimum mass, M_Psin i. However, we can calculate the expectation value of sin i for randomly oriented orbits to determine the most likely mass, and then use the giant planet mass–radius–flux relation derived by Weiss et al. (2013) to estimate the planetary radius:

$$\begin{equation} \frac{R_{\rm P}}{R_\oplus } = 2.45 \left(\frac{M_{\rm P}}{M_\oplus }\right)^{-0.039}\left(\frac{F}{{\rm erg \,s^{-1} \,cm^{-2}}}\right)^{0.094} \end{equation} \tag{ 2 }$$

where F is the time-averaged incident flux on the planet. Since R_P depends only weakly on M_P, assuming an inclination is not likely to introduce a large radius error—there is only a 1% difference in derived radius between edge-on and average-inclination orientations.

Under these assumptions, we find τ_cir ≈ 11.8 Gyr—much larger than the age of the cluster. Note that this holds true even for the full range of Q_Jup (corresponding to 700 Myr < τ_cir < 22.6 Gyr). We conclude that HD 285507b is dynamically young. While it is tempting to thus proclaim that migration has occurred via a HEM mechanism, recall that this is not a necessary condition for a dynamically young planet with an eccentric orbit. For any individual planet, non-zero eccentricity could also be the result of continued interaction with an undetected planetary or stellar companion, a recent close stellar encounter, or even modest eccentricity excitation via Type II migration (e.g., D'Angelo et al. 2006). Only analysis of a population of planets can provide meaningful insight into the migration process in this manner. Therefore, we turn to the literature for ages and circularization timescales of the known sample of hot Jupiters.

6.1. Description of the Analysis

To search for dynamical imprints of migration among known hot Jupiters (M_P > 0.3 M_Jup, P < 10 days), we follow the prescription described above to calculate τ_cir (and M_P and R_P for non-transiting planets). We adopt ages, eccentricities, planetary masses and radii, and stellar masses, radii, and temperatures from the literature.⁵ In Figure 4, we plot t_age versus τ_cir for this sample. While the figure is complicated by the uncertainties already discussed as well as poorly constrained ages and potential biases in measuring modest eccentricities (e.g., Shen & Turner 2008; Pont et al. 2011; Zakamska et al. 2011; Wang & Ford 2011), there is a hint that the points to the right of the circularization boundary are preferentially eccentric and the ones to the left are preferentially circular. If HEM were responsible for the final stages of hot Jupiter migration, this would be expected; planets get scattered inward on highly eccentric orbits and circularize over time. If Type II migration were responsible, we should expect very little difference between the eccentricity distributions to the right and left of the boundary; ordered migration through a gas disk should largely preserve circular orbits, so subsequent tidal interactions would not change the population significantly. In Figure 5, we plot the eccentricity histograms and cumulative distributions for the two populations, which contain 92 and 22 planets (as shown in Figure 4). To quantify the difference between them, we ran a Kolmogorov–Smirnov (KS) test. The KS p-value is the likelihood that the two subsamples came from the same parent distribution, and in this case, p = 3.3 × 10⁻⁵. We conclude that the two distributions do not come from the same parent distribution, with ∼99.997% confidence, and infer that high eccentricity migration mechanisms play a significant role in hot Jupiter migration. We also ran an Anderson–Darling (AD) test, which is similar to a KS test, but is more sensitive to differences in the distribution tails. The AD test indicates an even greater significance that the two distributions do not come from the same parent distribution of eccentricities.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

6.2. The Effect of Measurement Errors

Until now we have been assuming Q_P = 10⁶ to determine which planets are dynamically young (right side of the circularization boundary in Figure 4) and which are dynamically old (left side), allowing us to draw conclusions about the migration process. If we instead start with the assumption that planets migrate inward in possession of some initial eccentricity (rather than being gently shepherded on circular orbits through the gas disk), we can invert the problem to place a constraint on the tidal quality factor Q_P. As we vary Q_P, the circularization boundary changes location, and the difference between the two populations should be maximized (and the KS p-value minimized) for the correct (average) hot Jupiter Q_P. Note that it does not matter what fraction of hot Jupiters has undergone HEM—if any fraction has, then the minimum p-value should occur when the circularization boundary is in the correct place.

Figure 6 shows the results of this experiment. A value on the order of 10⁶ is preferred, which seems to rule out much of the parameter space consistent with Jupiter's tidal quality factor (and validates our assumption of Q_P = 10⁶). The discrete data points do not produce a smooth distribution, so finding the minimum is not straightforward. As such, we smooth it using a moving average with a boxcar filter of size [(2/3)Q_P, (3/2)Q_P]. Quantitative confidence limits on the minimum are difficult to assess (the vertical axis is not a probability associated directly with Q_P), but we take approximate upper and lower limits on Q_P to be those for which the smoothed p-value is 10 times its smoothed minimum value, which occurs for Q_P = 1.39 × 10⁶. For each simulated data set described in Section 6.2, we calculate Q_P this way and find the median to be similar (1.12 × 10⁶), but with smaller errors. This is not surprising, as the statistical errors from the simulations are akin to a standard deviation of the mean whereas the errors derived from a single p-value versus Q_P experiment are akin to a sample standard deviation. To describe the population, we therefore adopt the latter and suggest an appropriate tidal quality factor for a typical hot Jupiter is $\log {Q_{\rm P}} = 6.14^{+0.41}_{-0.25}$ (see Figure 6 for a visual representation).

Zoom In Zoom Out Reset image size

Our discovery of a hot Jupiter in the Hyades bolsters the statistics of short period giant planets in open clusters (of which three are now known). Including the Paulson et al. (2004) null result in the Hyades (0 planets among 74 FGK stars), our Hyades sample (1 in 26), and an updated census of Praesepe including unpublished data from our second year of observations (2 in 60), a total of 3 out of 160 stars host a hot Jupiter. After correcting for completeness and calculating Poisson errors following the prescription in Gehrels (1986), we find a hot Jupiter frequency of $1.97^{+1.92}_{-1.07}\%$ in the metal-rich Praesepe and Hyades open clusters. However, giant planet occurrence scales with metallicity approximately as 10^2[Fe/H] (Fischer & Valenti 2005). If we take [Fe/H] ≈ +0.15 as representative of the combined Praesepe and Hyades sample, the solar-metallicity-adjusted hot Jupiter frequency in clusters is $0.99^{+0.96}_{-0.54}\%$. Although more discoveries are needed to reduce the uncertainty, this is in good agreement with the frequency for field stars (1.20 ± 0.38%; Wright et al. 2012), and improves the evidence that planet frequency is the same in clusters and the field.

A primary motivation for the search for young planets is that their ages are comparable to the timescale of migration. Thus, the orbital properties of such planets may still bear the dynamical signature of this process. Since different migration mechanisms are predicted to produce hot Jupiters on different timescales and with different orbital eccentricities, we can use the properties of young hot Jupiters (and their existence at various ages) to determine the process by which they migrate. The ages of the cluster planets discovered thus far do not place a strong direct constraint on the timescale of migration (we know only that the process took less than 600 Myr), but the newly discovered planet in the Hyades holds a clue its dynamical history. HD 285507b has a long circularization timescale, so its non-zero eccentricity may be a remnant of the migration process, which would suggest planet-planet scattering or Kozai cycles have played a role in its orbital evolution. There is no observational evidence for a third body, but one cannot be excluded either. The RV timespan is not long enough to rule out a second giant planet (which also could have been ejected during scattering), and imaging by Patience et al. (1998) only rules out companions more massive than 0.13 M_☉ with projected separations 5–50 AU.

Applying this idea more broadly, we have compared ages and circularization timescales for all known hot Jupiters and find evidence for two families of planets, distinguished by their orbital properties: (1) mostly circular orbits for the "dynamically old" planets, those with τ_cir < t_age, and (2) a range of eccentricities for "dynamically young" planets, those with τ_cir > t_age. If Type II migration were the leading driver of hot Jupiter migration, both dynamically young and old planets should have circular orbits. We thus conclude that HEM is important for producing hot Jupiters. However, we can only say that these planets have experienced dynamical stirring at some point, and do not suggest that this evidence shows Type II migration to be unimportant. On the contrary, as shown by simulations time and again, Type II migration is almost certainly important to orbital evolution before the gas disk dissipates, but we suggest that for a large fraction of hot Jupiter systems, planet-planet scattering or the Kozai mechanism is responsible for the final stages of inward migration. A larger sample of dynamically young (non-circularized) planets may allow us to determine what that fraction is. Since few hot Jupiters have circularization timescales greater than 1 Gyr, a good way to enhance this sample is to continue finding young planets.

That HEM mechanisms play an important role in hot Jupiter migration has already been suggested, and is supported by a rich data set of stellar obliquity measurements in hot Jupiter systems (see Albrecht et al. 2012, for a recent discussion). In addition to the excitation of orbital eccentricity, dynamical encounters with a third body are expected to produce a range of orbital inclinations, although tidal interactions with the host star may realign the systems over time. These inclinations can be measured precisely, most notably via the Rossiter–McLaughlin effect (Rossiter 1924; McLaughlin 1924), and the results of such studies parallel those presented in this paper: systems for which the tidal timescale is short tend to be well-aligned, and those for which the timescale is long display high obliquities. Albrecht et al. (2012) do caution that stars and their disks may be primordially misaligned for reasons unrelated to hot Jupiters, but we see no obvious reason for this to influence the eccentricities. Migration through multi-body dynamical interactions, on the other hand, could explain both the inclined orbits and high eccentricities observed in systems that have not yet experienced significant tidal interactions. Whether that process is primarily planet-planet scattering or the Kozai effect remains to be determined, and it is likely that more data will be needed to properly answer this question.

Finally, the tidal circularization boundary that separates the dynamically young and old populations of hot Jupiters is sensitive to the choice of the planetary tidal quality factor, Q_P, so we have leveraged this dependence to constrain the typical value for hot Jupiters to be $\log {Q_{\rm P}} = 6.14^{+0.41}_{-0.25}$. Q_P has wide-ranging implications, e.g., for simulating orbital evolution (Beaugé & Nesvorný 2012) or modeling the inflated radii of hot Jupiters (Bodenheimer et al. 2003), but has thus far proven difficult to constrain observationally. While our result still includes substantial uncertainty and will not be applicable to any one planet, it can be applied to these problems in a statistical sense. Moreover, it offers a path forward: as our sample of longer period and young hot Jupiters grows, the determination of Q_P using this method should improve.

We thank Josh Winn, Tsevi Mazeh, and an anonymous referee for insightful comments and discussion, and Greg Feiden for producing a Dartmouth isochrone appropriate for our analysis. This research has made use of The Extrasolar Planets Encyclopaedia. The material herein is based upon work supported by the National Aeronautics and Space Administration (NASA) under grant No. NNX11AC32G issued through the Origins of Solar Systems program. S.N.Q. is supported by an NSF Graduate Research Fellowship, grant DGE-1051030. D.W.L. acknowledges partial support from NASA's Kepler mission under cooperative agreement NNX11AB99A with the Smithsonian Astrophysical Observatory. G.T. acknowledges partial support from NSF grant AST-1007992.

Facilities: FLWO:1.5m (TRES) - Fred Lawrence Whipple Observatory's 1.5 meter Telescope, FLWO:1.2m (KeplerCam) - Fred Lawrence Whipple Observatory's 1.2 meter Telescope