A Dwarf Planet Class Object in the 21:5 Resonance with Neptune

Dwarf planets in the trans-Neptunian region are remnant planetesimals from the protoplanetary disk of the solar system. They constrain the large-diameter end of the trans-Neptunian object (TNO) size distribution, which is inferred from the observed luminosity function (Brown 2008; Petit et al. 2008; Schwamb et al. 2013; Fraser et al. 2014). There is no simple size cut between dwarf planet and medium-sized TNO; whether an object achieves ellipsoidal hydrostatic equilibrium is dependent on its internal ice/rock fractional composition (Tancredi & Favre 2008; Lineweaver & Norman 2010). However, few suitably sizable worlds are yet known:¹³ 33 with, as a loose guide, absolute magnitude H_V < 4.

TNOs are faint due to their >30 au heliocentric distances, requiring discovery surveys by >1 m aperture wide-field optical telescopes. Past wide-area surveys have completed the inventory of bright TNOs to m_r ∼ 19.5 outside the galactic plane (Tombaugh 1961; Kowal 1989; Sheppard et al. 2000; Trujillo & Brown 2003; Moody 2004; Brown 2008; Brown et al. 2015). Substantial areas of sky have been surveyed to m_r ∼ 21.5 and deeper (Larsen et al. 2001, 2007; Elliot et al. 2005; Schwamb et al. 2010; Petit et al. 2011, 2017; Sheppard et al. 2011; Rabinowitz et al. 2012; Sheppard & Trujillo 2016; Gerdes et al. 2017).

As often the brightest and thus easiest to detect of the worlds in the trans-Neptunian region, dwarf planets also provide a useful broad-brush indication of the phase space of their dynamical populations. They are key to exploring the fainter, large-semimajor-axis populations where the smaller TNOs are too faint to detect; for example, the 2003 discovery of (90377) Sedna indicated a substantial population of TNOs with large perihelion distances (Brown et al. 2004). The a ≳ 50 au populations are all defined by their degree of gravitational interaction with Neptune: they include orbits librating in high-order mean-motion resonance (MMR); the scattering disk, on orbits actively interacting with Neptune; and the "detached" TNOs, with perihelia q ≳ 37 au (Gladman et al. 2008). Recent discoveries include the 9:2 mean-motion resonant object 2015 RR₂₄₅ with H_r = 3.6 and a = 81.86 ± 0.05 au (Bannister et al. 2016a), the scattering disk TNO 2013 FY₂₇ with H_V = 2.9 and a = 59 au (Sheppard & Trujillo 2016), and the detached TNO 2014 UZ₂₂₄ with H_V = 3.5 and a = 109 ± 7 au (Gerdes et al. 2017).

The Panoramic Survey Telescope and Rapid Response System 1 Survey (Pan-STARRS 1, hereafter PS1) is well suited to the discovery of TNOs. PS1 is a 1.8 m telescope on Haleakela in Hawai'i, with a dedicated 0258 pixel, 7 deg² optical imager (Kaiser et al. 2010; Chambers et al. 2016). The PS1 3π survey (Magnier et al. 2013; Chambers et al. 2016) repeatedly observed the sky north of decl. −30° using a Sloan-like filter system (Tonry et al. 2012), reaching typical single-exposure 5σ depths of g_P1 = 22.0, r_P1 = 21.8, and i_P1 = 21.5 (see Table 11 of Chambers et al. 2016). PS1 also observed within ±20° of the ecliptic using the wide w_P1 filter (m_5σ ∼ 22.5). The observing cadence of PS1 is optimized for detection of inner solar system minor planets, with analysis for detection made with the PS Moving Object Processing System (MOPS; Denneau et al. 2013). However, PS1's many visits permit detection of slower-moving (∼3''/hr) TNOs in the accumulated data. Weryk et al. (2016) reported several hundred centaurs and TNOs from a search of the 2010 February 24 to 2015 July 31 PS1 observations, which linked together detections within 60 day intervals.

The PS1 outer solar system (OSS) key project uses a novel linking solution, based on transforming topocentric observations to a heliocentric coordinate frame using an assumed heliocentric distance. Our initial search of the 2010–2014 PS1 observations resulted in hundreds of candidates, ∼50% of which were newly discovered TNOs (Holman et al. 2015). These include unusual objects such as the highly inclined centaur (471325) 2011 KT₁₉ (Chen et al. 2016), as well as numerous new Neptune trojans (Lin et al. 2016).

Here, we report the discovery of an m_r ∼ 21 dwarf planet candidate at a barycentric distance of 55 au: 2010 JO₁₇₉. We present the technique used to detect 2010 JO₁₇₉ (Section 2), our observations (Section 3), 2010 JO₁₇₉'s physical properties (Section 4), the dynamical classification of its orbit (Section 5), and the broader implications of its existence for our understanding of the distant, dynamically excited TNO populations (Section 6).

We developed our discovery pipeline to cope with the temporal sparsity of the Pan-STARRS data (see Brown et al. 2015 for an independent, alternative approach to detecting TNOs in sparse data sets). Our pipeline operates on the catalogs of source detections found in the direct, undifferenced PS1 exposures spanning 2010 to mid-2014 by the Image Processing Pipeline (IPP; Magnier 2006, 2007). We eliminate any detection that coincides within 1'' of a known stationary source. (A catalog of stationary sources was developed from the individual detections in the PS1 exposures; detections that occur near the same location over multiple nights are considered to be stationary.) The remaining transient detections form the input to the rest of the pipeline (we do not use MOPS). All magnitudes are transformed to w-band (Tonry et al. 2012) for uniformity of comparison. We iterate over a set of heliocentric distances, d ∼ 25–1500 au. For each assumed distance, we carry out a number of steps. First, we transform the topocentric sky plane positions of transient sources we identify in the PS1 imaging to those as would be observed from the Sun. Then, all tracklets are identified: these are sets of ≥2 detections from the exposures within each individual night that are consistent with linear motion that would be bound to the Sun. No minimum rate of motion is required. Tracklets with three or more detections are much more likely to be real, as the positions of their constituent detections must be consistent with linear motion at a constant rate. For every pair of high-confidence tracklets (≥3 detections) that can be associated with a bound heliocentric orbit, we look for additional supporting tracklets along the great circle defined by those two tracklets. If an additional tracklet is found, an orbit is fit (using modified routines from the Orbfit package of Bernstein & Khushalani 2000) to the set of observations from those three tracklets, and a search is carried out for additional tracklets along the sky plane trajectory defined by that orbit. As tracklets are found, the orbit is refined and the search continues, recursively. All of the different linking possibilities are followed until the set of plausibly connected tracklets is exhausted. As a final step, single detections that lie within 10 of the sky plane are searched for and incorporated, with astrometric and photometric outliers rejected.

We consider for further investigation any arc of tracklets with detections on at least five separate nights that has an orbital solution with a reduced chi-squared ${\bar{\chi }}^{2}\lesssim 3$ and that has a range of observation magnitudes that is physically realistic (Δm_w < 1.5). All directions of motion of the resulting orbit are permitted and retained.

The observations of 2010 JO₁₇₉ span 12 oppositions. All available photometry is tabulated in the Appendix (Table 3); the astrometry is listed at the Minor Planet Center.¹⁴

We initially detected 2010 JO₁₇₉ in g-, r-, and i-band observations spanning 2010–2012 from the PS1 3π survey (Table 3). The absence of w-band observations is due to 2010 JO₁₇₉'s 32° ecliptic latitude at the time of discovery, outside the coverage of the PS1 w-band survey. 2010 JO₁₇₉ is seen on 12 distinct nights with a total of 24 detections, forming an arc spanning 790 days. 2010 JO₁₇₉ was retained as a candidate TNO as it passed two tests: (a) the residuals to an orbital solution determined with a modified version code of Bernstein & Khushalani (2000) were consistent with the astrometric uncertainties of the individual detections (∼01), with no outliers; (b) the photometric measurements from the PS1 IPP, transformed to w-band (Tonry et al. 2012), spanned only 1.2 mag. We visually examined the detections in the PS1 images for final verification that each was genuine. The photometry in Table 3 is calibrated PS1 Data Release 1 (PS1-DR1; Magnier et al. 2016) and was measured with the moving-object photometry analysis package TRIPPy (Fraser et al. 2016).

Although 2010 JO₁₇₉ is substantially brighter than the PS1 detection limits, there is a significant bias against finding such objects with the algorithm we described in Section 2. First, we note that only two tracklets with three or more detections can be seen in Table 3. This is the minimum number for our algorithm. If either of these tracklets was removed from the data set, the algorithm would not have found 2010 JO₁₇₉. This could occur moderately often, given the 76% fill factor of the focal plane (Chambers et al. 2016). Interestingly, the later of those two tracklets was accidental: it was found in the overlap between two adjacent fields for which pairs of observations were being taken. The overlap region is 10%–20% of the survey area, depending upon the specific survey pattern.

We followed up 2010 JO₁₇₉ with Sloan r-band observations on 2016 July 28–30 with the EFOSC2 camera on the New Technology Telescope (NTT; Buzzoni et al. 1984; Snodgrass et al. 2008) atop La Silla, Chile. EFOSC2 uses a LORAL 2048 × 2048 CCD that was used in 2 × 2 binning mode. Each binned pixel maps onto an on-sky square 024 on the side, for a full field of view 41 × 41. The data were subject to standard bias-subtraction and flat-fielding procedures, and the magnitude of 2010 JO₁₇₉ in each frame was measured using circular aperture photometry and calibrated to PS1-DR1 using tens of field stars.

2010 JO₁₇₉ is bright enough to be found in archival images with the SSOIS search tool (Gwyn et al. 2012). We recovered astrometry of 2010 JO₁₇₉ from the griz observations of a sequence of 54 s exposures (riuzg) made on 2005 May 11. The observations by the 2.5 m telescope of the Sloan Digital Sky Survey (SDSS) in New Mexico were part of SDSS-II (Abazajian et al. 2009). The astrometry of the 0396 pixel images is calibrated to SDSS Data Release 14 (Abolfathi et al. 2017). The TNO was close by stars, which prevented photometric measurements.

We also found via SSOIS that 2010 JO₁₇₉ was serendipitously imaged in Sloan g, r, and z on 2014 August 16–18 by the DECalS survey¹⁵ with the Dark Energy Camera (DECam; DePoy et al. 2008) on the 4 m Blanco Telescope in Chile. DECam has 0263 pixels with a 3 deg² field of view. The DECalS astrometry and photometry zeropoints are calibrated¹⁶ to PS1-DR1. The five highest signal-to-noise ratio (S/N) and best seeing DECam images are our overall highest-S/N images of 2010 JO179. Photometry, point-spread function (PSF) modeling, and trailed PSF removal were performed with TRIPPy (Table 3). No evidence of binarity was found, to an estimated signal-to-noise ratio (S/N) = 5 detection threshold and a separation of ∼1 FWHM.

The mean colors of 2010 JO₁₇₉ were calculated from the mean magnitudes in each band, after correction to unit heliocentric and geocentric distance (Figure 1). Standard deviations for each band were combined in quadrature to yield the uncertainty. Table 1 summarizes the derived color measurements. The substantial time between observations means rotation could have let potential surface variability intrude (Fraser et al. 2015; Peixinho et al. 2015), so we avoid presenting the colors as a coarsely sampled spectrum. The g − r and r − i of 2010 JO₁₇₉ fall along the locus of the known range of TNOs (e.g., Ofek 2012), classing it as moderately red.¹⁷

Zoom In Zoom Out Reset image size

The visual albedos constrained by thermal measurements for 2 < H_r < 4 TNOs are wide-ranging, varying from p = 0.07–0.21 (Brucker et al. 2009; Lellouch et al. 2013; Fraser et al. 2014). As the color-albedo clustering seen for smaller TNOs does not continue to objects this large (Lacerda et al. 2014; Fraser et al. 2014), the albedo of 2010 JO179 is unconstrained within the known albedo range. At H_r = 3.44 ± 0.10 (see below), 2010 JO₁₇₉ has a diameter of at least 600 km; at the less reflective albedo limit, it could be as large as 900 km. It will thus achieve ellipsoidal hydrostatic equilibrium (Tancredi & Favre 2008; Lineweaver & Norman 2010).

2010 JO₁₇₉'s r − z color is unusually blue. To make it consistent with the range of r − z colors exhibited by other TNOs with g − r colors similar to 2010 JO₁₇₉ (Pike et al. 2017a) requires a 1σ deviation in both g − r and r − z, or a 2σ deviation in either. Yet the mean r − z is consistent with the only single-epoch r − z measurement (from DECam; see Table 1). Alternatively, the negative r − z color is broadly consistent with the presence of methane ice, which exhibits an absorption feature through the z band (see, for example, Tegler et al. 2007). For 2010 JO₁₇₉ to have a methane-bearing surface would be unusual, as it is expected that objects of this size would rapidly lose that volatile to space on short timescales (Schaller & Brown 2007b), in contrast to the larger dwarf planets, e.g., (50000) Quaoar and 2007 OR₁₀ are red methane-covered objects (Schaller & Brown 2007a; Brown et al. 2011).

We use the photometry in Table 3 to constrain the color, phase curve, and light curve properties for 2010 JO₁₇₉. The apparent magnitudes were first corrected to units of heliocentric (r) and geocentric (Δ) distance by subtraction of $5\mathrm{log}(r{\rm{\Delta }})$. The corrected g and r magnitudes were fit simultaneously with the following model, which accounts for a linear phase darkening and sinusoidal light curve variation:

$$\begin{eqnarray}&&{H}_{g,r}-(g-r)+\beta \alpha +\displaystyle \frac{{\rm{\Delta }}m}{2}\sin \left[\displaystyle \frac{2\pi (t-{t}_{0})}{P}\right],\end{eqnarray} \tag{ 1 }$$

where H_g,r is the absolute g or r magnitude; α is the phase angle; β is the linear phase function slope; Δm, P, and t₀ are the peak-to-peak variation, period, and offset of the light curve, respectively; and (g − r) is a color term subtracted from the g magnitudes only. As is the case for Earth-based TNO observations, the range of phase angle is limited: 0 < α ≲ 1°. We found the best-fit β to be somewhat dependent on the initialization values for P, Δm, and β. To solve this, we varied (g − r) linearly in steps of 0.01 mag and at each step generated 200 uniformly distributed random initial values $P\in (16,64)$, ${\rm{\Delta }}m\in (0.15,0.60)$, and $\beta \in (0.01,0.50)$. The range of β values brackets the slopes seen in TNOs (Rabinowitz et al. 2007). The NTT r-band photometry displays a steady brightening by about 0.2 mag over a period of 5 hr, which provides useful limits on the light curve variation (Δm > 0.2 mag) and period (P > 10 hr). We found consistently that shallow phase functions (β < 0.05 mag deg⁻¹) fit the data best (see Figure 2). For this reason we decided not to correct the data for phase effects before the light curve analysis.

Zoom In Zoom Out Reset image size

We employed the Lomb–Scargle algorithm (Press & Rybicki 1989) to measure the light curve period using the r-band measurements only. The periodogram identified the period P₁ = 30.6324 hr as the best solution, which corresponds to a single-peaked light curve, with one maximum and one minimum per rotation. Figure 3 shows the data phased with P₁. The double period P₂ = 2P₁ = 61.2649 hr is also shown for comparison, in case 2010 JO₁₇₉ has a symmetric light curve.

Zoom In Zoom Out Reset image size

Both the solutions are consistent with the data and indicate that 2010 JO₁₇₉ is a slow rotator. The single-peaked solution, P ∼ 30.6 hr, would imply that 2010 JO₁₇₉ is roughly spherical and has significant albedo patchiness. The double-peaked solution would imply an ellipsoidal shape with axes ratio a/b > 1.58. We find this less plausible, given the large size and slow rotation of 2010 JO₁₇₉.

Finally, the light curve period solutions were used to refine the phase curve, resulting in the linear solution shown in Figure 2, with β = 0.02 mag deg⁻¹. We force β > 0 in our fitting algorithm, which sets the lower limit on the phase function slope, and find a 1σ upper limit β < 0.07 deg mag⁻¹. We find an absolute r magnitude H_r = 3.44 ± 0.10 and a best-fit color offset g − r = 0.85, consistent with the g − r values in Table 1.

We determine the barycentric elements of 2010 JO₁₇₉ by fitting observations from six oppositions spanning 2005–2016 (Table 2), using the code of Bernstein & Khushalani (2000). The barycentric distance to 2010 JO₁₇₉ at discovery in 2010 is 55.019 ± 0.003 au.

We generate 100 clones of the orbit of 2010 JO₁₇₉, varying the observations by their respective uncertainties and refitting the orbit, hence creating a statistically rigorous bundle of orbits that are all consistent with the observations. We integrate the best-fit orbit and clones as test particles in a barycentric system, orbiting in the gravitational field of the Sun and four giant planets. We integrate the particles using the adaptive IAS15 integrator (Rein & Spiegel 2015) in the REBOUND code of Rein & Liu (2012) for 700 Myr, i.e., around 10⁶ orbits of 2010 JO₁₇₉.

The period ratios with Neptune, eccentricity and inclination for the best-fit orbit (black), and a sample of the clones (all other colors) are shown in the top three panels of Figure 4. All of the particles stably orbit with period ratios very close to 4.2. We then consider the resonant angle,

$$\begin{eqnarray}&&{\theta }_{p+q:q}=(p+q){\lambda }_{\mathrm{out}}-p{\lambda }_{\mathrm{in}}-q{\varpi }_{\mathrm{out}},\end{eqnarray} \tag{ 2 }$$

and plot in the lower panel of Figure 4 values for Equation (2) using p + q:p = 21:5. The best-fit orbit exhibits stable libration in the 21:5 resonance for ∼100 Myr before diffusing to a circulating configuration with a period ratio ∼4.21 with respect to Neptune for ∼300 Myr, before returning to a 21:5 resonant configuration for the remainder of the simulation. We note that 16 possible resonant angles exist for the 21:5 resonance in which the −qϖ_out term in Equation (2) is replaced by −(nϖ_in + (q − n)ϖ_out), with 0 ≤ n ≤ 15. We examined all such variations for all best-fit and clone orbits, and found that only the n = 0 case (i.e., Equation (2)) ever exhibits resonance.

Zoom In Zoom Out Reset image size

The clones of the orbit of 2010 JO₁₇₉ display behavior consistent with that of the best-fit orbit. For the entirety of the simulation, all clones have period ratios with Neptune that remain bounded between 4.185 and 4.215 (semimajor axes bounded between ∼78.1 au and ∼78.5 au) and have pericenters in the range 37.5–41.5 au. The narrowness of the 21:5 resonance means even the small remaining orbital uncertainties encompassed by the clones are of similar scale to structures in the resonance's phase space. Most clones move back and forth between resonant and non-resonant configurations, with a slow overall diffusion away from resonance. All clones are resonant at the start of the simulation, and ∼6% remain resonant for the entire simulation, so there is some possibility that 2010 JO₁₇₉ is on one of these stable orbits. The median time at which clones exit resonance is ∼110 Myr. At 700 Myr, ∼25% of clones are resonant.

From the majority behavior of the best-fit orbit and its clones over the 700 Myr simulation, 2010 JO₁₇₉ probably has a metastable orbit, which switches between resonant and non-resonant configurations on hundred-Myr timescales.

Although 2010 JO₁₇₉ is bright for a TNO at m_r ∼ 21, its 32° orbital inclination (Figure 5, right) and resulting current ∼30° ecliptic latitude is responsible for it not being detected earlier. It is bright enough to have been detected by the surveys of Larsen et al. (2007) and Schwamb et al. (2010), but fell just outside their sky coverage. It was not detected by Weryk et al. (2016), presumably either because 2010 JO₁₇₉ is outside of the region considered in that search, or it was not observed with their required cadence. 2010 JO₁₇₉ is thus a good example of how the detection efficiency of a survey is a function of each specific analysis, given a common observational data set. Surveys that cover large fractions of the sky, both well away from the ecliptic and to fainter limiting magnitudes, such as are planned for the forthcoming Large Synoptic Survey Telescope (LSST; Science Collaboration et al. 2009), will discover many more such objects. However, archival data will clearly still yield new discoveries with further thorough searching.

For population studies, the detection efficiency of surveys such as PS1 need to be well characterized, as has been done for the CFEPS and OSSOS surveys (Petit et al. 2011; Bannister et al. 2016b), to be able to correct for the observational biases. In addition, detailed dynamical classifications of the discoveries, of the type we have done here, need to be made. These permit determination of the intrinsic abundance of TNOs such as 2010 JO₁₇₉ (e.g., Pike et al. 2015). This is the subject of future work with PS1.

With a perihelion distance q = 39.32 au (Figure 5), near the boundary between what are considered to be low perihelion and high perihelion TNOs (Lykawka & Mukai 2007a; Gladman et al. 2008), 2010 JO₁₇₉ is a relatively nearby example of what is now being recognized as a very substantial population of high-perihelion TNOs (Gladman et al. 2002; Gomes 2003; Trujillo & Sheppard 2014; Pike et al. 2015; Kaib & Sheppard 2016; Nesvorný & Vokrouhlický 2016; Nesvorný et al. 2016). In the absence of a survey characterization, we cannot provide an absolute estimate for the 21:5 resonant population, especially given the strong observability bias on the eccentricity distribution of such a large-a population. However, we can find a general lower limit. The existence of 2010 JO₁₇₉ requires there to be at least one H_r ≃ 3.5 TNO. Scaling according to the size distribution of the dynamically excited TNOs (Fraser et al. 2014), there are at least 6700 objects in the 21:5 that are larger than 100 km in diameter. This would be more numerous than the 3:2 plutinos (Gladman et al. 2012); note the sizable known sample of this population visible in Figure 5, and is consistent with the large populations found in other distant resonances (Pike et al. 2015; Volk et al. 2016). We note that the 21:5 is not an intrinsically "special" resonance: this argument applies more generally for any large-e orbit with q ∼ 39 au (e.g. Bannister et al. 2016a), and the 21:5 occupancy only reinforces the vast scale of the populations in this region. Aspects of these large populations remain challenging to form in migration scenarios (cf. Pike et al. 2017b).

Zoom In Zoom Out Reset image size

Neptune's early migration into the outer planetesimal disk plausibly emplaced the dynamically excited trans-Neptunian populations, including the resonant and scattering disk (Malhotra 1995; Gomes 2003; Gladman et al. 2008). Predicting the details of the emplaced distant populations is an area of active investigation. Nesvorný et al. (2016) and Kaib & Sheppard (2016) independently predict that some TNOs with a > 50 au and q > 40 au should exhibit semimajor axes clustered near and inward, but not within, MMRs with Neptune. This followed from modeling Neptune's dynamical evolution on an orbit with moderate e ≲ 0.1 eccentricity, both under smooth migration and from "grainy" gravitational interaction with a small sea of dwarf planets in the initial planetesimal disk (Nesvorný & Vokrouhlický 2016). In direct contrast, Pike & Lawler (2017) predict fairly symmetric trails of high-perihelia populations immediately surrounding distant MMRs. They assessed emplacement by the Nice model scenario of Brasser & Morbidelli (2013), with a smoothly migrating, initially high e = 0.3 Neptune. While one might hope to constrain the details of migration from the present-day orbit of a distant TNO like 2010 JO₁₇₉, it unfortunately lies in a region of phase space where its orbit does not distinguish among current model outcomes. Irrespective of whether Neptune's migration was smooth or grainy, fast or slow, the results of Nesvorný et al. (2016), Kaib & Sheppard (2016), and Pike & Lawler (2017) all contain TNOs with a ∼ 78 au, q ∼ 39 au, i ∼ 30°.

It is also worth considering if 2010 JO₁₇₉ could be a more recent arrival to its current orbit. The a > 50 au region is filigreed with high-order resonances. These permit "resonance sticking," where a TNO's orbit temporarily librates in resonance for tens to hundreds of millions of years, chaotically escapes the resonance, changing in semimajor axis, then sticks, librating within another resonance (e.g., Lykawka & Mukai 2007b). The resonant dynamics during a temporary capture can lead to oscillations in eccentricity and inclination that can then weaken the TNO's interaction with Neptune (Gomes et al. 2008; Sheppard et al. 2016). However, 2010 JO₁₇₉'s clones all remain close to its resonance (Section 5), implying stability for the age of the solar system, rather than showing behavior like that of 2015 RR₂₄₅, which moves between the 9:2 resonance and the scattering disk (Bannister et al. 2016a). 2010 JO₁₇₉'s clones exhibit similar behavior to the four known TNOs in and by the 5:1 resonance at a ∼ 88 au (Pike et al. 2015). The current orbit of 2010 JO₁₇₉ is therefore more likely to be ancient.

M.J.H. and M.J.P. gratefully acknowledge NASA Origins of Solar Systems Program grant NNX13A124G and grant NNX10AH40G via sub-award agreement 1312645088477, BSF Grant 2012384, NASA Solar System Observations grant NNX16AD69G, as well as support from the Smithsonian 2015 CGPS/Pell Grant program. M.T.B. appreciates support from UK STFC grant ST/L000709/1.

The Pan-STARRS 1 Surveys (PS1) have been made possible through contributions of the Institute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its participating institutes, the Max Planck Institute for Astronomy, Heidelberg and the Max Planck Institute for Extraterrestrial Physics, Garching, The Johns Hopkins University, Durham University, the University of Edinburgh, Queen's University Belfast, the Harvard-Smithsonian Center for Astrophysics, the Las Cumbres Observatory Global Telescope Network Incorporated, the National Central University of Taiwan, the Space Telescope Science Institute, the National Aeronautics and Space Administration under grant No. NNX08AR22G issued through the Planetary Science Division of the NASA Science Mission Directorate, the National Science Foundation under grant No. AST-1238877, the University of Maryland, and Eotvos Lorand University (ELTE).

This work is based in part on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere under ESO programmes 194.C-0207(H), and on observations at Cerro Tololo Inter-American Observatory, National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation.

The computations in this Letter were run on the Odyssey cluster supported by the FAS Science Division Research Computing Group at Harvard University.

This research used the facilities of the Canadian Astronomy Data Centre operated by the National Research Council of Canada with the support of the Canadian Space Agency.

Facilities: PS1 - Panoramic Survey Telescope and Rapid Response System Telescope #1 (Pan-STARRS), NTT (EFOSC2) - , Sloan - , CTIO:Blanco (DECam) - .

Software: Astropy, TRIPPy, REBOUND, Matplotlib.

The Appendix comprises Table 3.