OSSOS. VII. 800+ Trans-Neptunian Objects—The Complete Data Release

The 13BL block was a 3 × 7 grid of MegaCam fields that overlaid the invariable plane, centered at 055', 4°00' (Figure 2). The 13BL discovery observations were made in median IQ of 075 with the R.MP9601 filter. Half the block was acquired on 2013 September 29, and the other half was acquired on 2013 October 31. A complete triplet sequence was acquired on 20 MegaCam fields (one fewer than in the 21-field block design). Uranus was about 1° away during discovery acquisition but did not contribute any scattered light. Tracking observations within the discovery lunation and in the lunations on either side were smoothly acquired. Pointed tracking observations were made from 2014 July through 2015 January, all with the R.MP9601 filter.

The 14BH block was a 3 × 7 grid of pointings 2°–5° off the invariant plane, centered at 129', 12°58' (Figure 2). Poor weather during the 2013 October opposition prevented observation of a valid discovery triplet for the whole H block. Only six fields received a survey-quality m_r ∼ 24.5 depth triplet on 2013 November 1. Fortunately, the multiyear nature of the Large Program allowed the discovery observations for the H block to be deferred to the next year's opposition. Also, the CFHT mirror was re-aluminized in 2014 July, a process that increases optical throughput. The full 21 fields of 14BH received discovery triplet observations on a single night, 2014 October 22, in the R.MP9601 filter. All discoveries at Kuiper Belt distances found in the six-field 2013 triplet were rediscovered in the 2014 triplet. The capriciousness of weather meant the designed OSSOS survey cadence was modified for this block: there were two semesters of only tracking the 14BH grid, with the discovery triplet observed in the second. The cadence and strategy remained successful under this strain. During analysis, the density of the 2013 observations permitted immediate linking out of each TNO's orbit from the arcs in the 2014 discovery semester to "precovery" arcs throughout 2013B. Despite the unorthodox observing, only minimal cleanup observations with pointed ephemerides were needed in 2014 December and 2015 January. These secured the orbits of a small fraction of <30 au discoveries, which had moved to areas of sky outside the grid of the block that was observed in 2013–2014.

The 15AP block was a 4 × 5 grid of pointings spanning the invariant plane and centered at 1330', −7°45' (Figure 2). Its positioning, lower in R.A. than 13AE, centers it more on the invariant plane while avoiding potential scattered light from α Vir. The 15AP block was acquired for discovery on 2015 April 12 in 06 seeing in the R.MP9602 filter (discussed below).

A new set of filters was introduced for CFHT's MegaPrime in 2015. The new filters allowed the use of four chips that have been on MegaCam since its commissioning but were vignetted out. This changed the MegaCam focal plane from 36 chips to 40 chips, which required rearranging the OSSOS grid to a 4 × 5 tessellation. All 2015A observations, including the discovery triplets, were acquired in MegaCam's new R.MP9602 filter. Section 2.8.3 discusses how R.MP9602 has a very similar bandpass but 0.13 mag higher throughput than the R.MP9601 used in previous years (Figure 3). Use of R.MP9602 maintained consistency in the discovery bandpass across the whole of OSSOS. Section 4 discusses how TNO colors can thus be modeled.

The new wide-band filter for MegaCam, GRI.MP9605, spans 4000–8200 Å and is very flat in transmission. We discuss its characteristics in Section 2.8 (Figure 3). Observations in GRI.MP9605 are referred to as w in our observations reported to the MPC. We used the GRI.MP9605 filter exclusively for all of our tracking observations after it became available in 2015 August: it added an extra ∼0.7 mag of depth (Section 2.8.3), ideal for recovery of fainter TNOs near the characterization limit.

Good weather in the 15A semester allowed pointed recoveries in GRI.MP9605 on the TNOs discovered in 15AP to proceed without incident from 2016 January through late July. Two observations, spaced by a week to 10 days, were made each lunation.

The off-plane 15AM block was a 4 × 5 grid of pointings nestled next to the earlier 13AO block (Bannister et al. 2016b) at a slightly lower ecliptic latitude, centered at 1530', −12°20' (Figure 2). The multiyear nature of our Large Program permitted advance observations: we predicted the 2014 location of model Kuiper Belt objects that would be within the desired region of sky at opposition in 2015. In 2014, we imaged 21 deg² (arranged in the 36 chip 3 × 7 grid layout) with a triplet in R.MP9601, with half on 2014 May 29 in 05–07 seeing and the remainder on 2014 June 1 in similar IQ. In 2015, a successful discovery triplet in the new R.MP9602 filter (see Section 2.3) saw 05 IQ on each half, observed on 2015 May 24 and 25. Despite the change in the grid layout (see Section 2.3) and the comparatively higher dispersal rate on the sky of TNOs on more inclined and eccentric orbits, the 2014 observations were able to provide precovery for more than half of the 15AM discoveries. Tracking throughout 16A in the new GRI.MP9605 filter proceeded smoothly (see Section 2.3).

The 15B observations provide the deepest areas of sky observed by OSSOS, with discovery observations in MegaCam's new filters R.MP9602 and tracking observations in GRI.MP9605 (see Section 2.3) in a semester featuring exceptionally good weather.

The 15BS block was a 4 × 5 grid of pointings placed above the invariant plane to sample inclinations midway between those observed by 13BL and 14BH and centered at 030', +05°00' (Figure 2). Its discovery triplet acquisition in R.MP9602 in 2015B was in two halves, split parallel to the plane. Both were observed in excellent seeing. However, the half observed on 2015 September 9 had better IQ than that on 2015 September 8, and the half acquired on September 8 had a small pointing shift error for one image of the triplet. For characterization, we thus regard it as two adjacent blocks of equal size: 15BS, the southern half, shallower and closer to the invariant plane, and 15BT, deeper and higher-latitude, as shown in Figure 2.

The tracking acquisitions in GRI.MP9605 in 2015B proceeded smoothly. However, 2016B saw several dark runs become near-total losses due to weather. A quarter of the 2015B discoveries from 15BS/15BT received insufficient tracking for high-precision orbit determination and were targeted for additional 2017 observation.

The 15BD block was a 4 × 5 grid of pointings centered on the invariant plane at 0315', +16°30' (Figure 2). As with the 15AM block (Section 2.4), advance planning allowed acquisition of precovery imaging in 2014. This included acquisition on 2014 November 17 of a lengthened, 400 s exposure triplet in R.MP9601 on 11 deg² of the block. Designed to be the deepest block OSSOS would acquire, the 2015 discovery triplet's exposure times were lengthened to 400 s in R.MP9602. We redesigned the acquisition cadence to ensure that a triplet spanning 2 hr could still be successfully acquired on a large enough area of sky. Notable variation in IQ occurred between the two nights of observation: 06 IQ on 2015 November 6 and 045 IQ on 2015 November 7. The characterization on this block is therefore also split in two, forming the adjacent 15BC and 15BD areas (Figure 2). Both have the same latitude coverage. The shallower 15BC is in two parts: a 1 × 4 deg² column and a 2 × 4 deg² area, as shown in Figure 2. These bracket the contiguous 2 × 4 deg² area of 15BD, which is the deepest area imaged by OSSOS.

Similar to that for block 15BS (see Section 2.5), the tracking observations in GRI.MP9605 in 15B were well sampled. The 15BD block was buffered against the sparse observing that occurred in 2016B (see Section 2.5) by the partial-coverage observations in 2014. Exposure times for the GRI.MP9605 tracking observations were increased to 450 s to retain the deep discoveries.

After 2 yr of intensive observation, a small fraction of TNOs either had orbits still classifying them as insecure or were securely resonant but had large libration amplitude uncertainties. These were targeted with a few extra observations in GRI.MP9605: at minimum two, and in some cases three or four epochs, at times far from opposition that provided the greatest parallax. In 2017, these observations were provided via a Director's Discretionary (17A) and regular-time (17B) proposal. The CFHT mirror was re-aluminized in 2017 August, again producing slightly deeper images.

Additional observations came from a Director's Discretionary program with CFHT MegaCam in 2014–2017, which simultaneously observed a magnitude-limited m_r < 23.6 sample of OSSOS TNOs together with Gemini North for Colours of OSSOS (Col-OSSOS; e.g., Fraser et al. 2015). The program observed a uniform r–u–r filter sequence, with the two 300 s r-band images separated by 1.5 hr. The r filter was R.MP9601 in 2014B-2015A and R.MP9602 after 2015 August. The Col-OSSOS imaging covered most of the area of 13AE, 13AO, 13BL, 14BH, 15BS, and 15BT (M. T. Bannister et al. 2017, in preparation). Its r-band observations are folded into the OSSOS analysis.

The targeted imaging and that for Col-OSSOS produced substantial amounts of serendipitous observing of the TNOs nearby on the sky due to the large MegaCam field of view. The orbits (Section 3) in this paper thus update those given in Bannister et al. (2016b) for the discoveries from blocks 13AE and 13AO.

OSSOS required high-precision, coherent astrometric calibration to produce TNO orbits of the desired quality (Section 3.3). During the 2013 February–2016 October tracking of each TNO, we acquired observations and then reclassified the newly improved orbit to determine if additional observations were required to secure the orbit. The image calibration during this time used an iterative approach with the 2MASS and UCAC4 catalogs, described in Bannister et al. (2016a, Section 3). Since the Gaia Data Release 1 was published in late 2016, all OSSOS data have been astrometrically calibrated using Gaia. At the end of the Large Program observations, all earlier OSSOS images were recalibrated using Gaia. The orbits that we report here (Section 3) and to the MPC are based on this calibration.

2.8.1. Astrometric Calibration

2.8.2. The Effects of Differential Chromatic Refraction

2.8.3. Photometric Calibration

The filter responses for the three OSSOS filters are shown in Figure 3. These filter responses include the full system of the telescope and camera (including mirror, lenses, and the quantum efficiency of the detector), 1.25 airmasses of atmosphere, and the transmission of the filters themselves. The newer r-band filter (R.MP9602) is slightly wider than the older r-band filter (R.MP9601) and has higher transparency over a wider wavelength range. Images taken in R.MP9602 are on average 0.13 mag deeper than those taken with R.MP9601. Images taken in the wide filter (GRI.MP9605) are on average 0.71 mag deeper than those taken in R.MP9601.

Previously, the photometric calibration was based on the SDSS (Fukugita et al. 1996; Ahn et al. 2014). As described in Bannister et al. (2016a, Section 3), images taken on the footprint of the SDSS were directly calibrated using the SDSS catalog converted into the MegaCam system. For images off the SDSS footprint, the procedure was more intricate. First, we computed the zero-point variation across the focal plane of MegaCam. Then, for each photometric night, all images on the SDSS were used to compute a nightly zero point. Images taken on photometric nights were calibrated using this zero point. Images taken on nonphotometric nights were calibrated by transferring the zero point from images taken on photometric nights using a bootstrapping system.

With the release of PS-DR1, this system was replaced with a much simpler system. Each image was calibrated using the PS-DR1 photometric catalog. The Pan-STARRS photometry was converted to the MegaCam photometric system,

$$\begin{eqnarray}{\rm{R}}.\mathrm{MP}9601 & = & r+0.002-0.017x+0.0055{x}^{2}\\ & & -0.00069{x}^{3},\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}{\rm{R}}.\mathrm{MP}9602 & = & r+0.003-0.050x+0.0125{x}^{2}\\ & & -0.00699{x}^{3},\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}\mathrm{GRI}.\mathrm{MP}9605 & = & r+0.005+0.244x-0.0692{x}^{2}\\ & & -0.00140{x}^{3},\end{eqnarray} \tag{ 3 }$$

where r refers to r_Pan–STARRS and x = g − i in the Pan-STARRS system. This third-order polynomial expression of color terms in g − i follows the convention in Finkbeiner et al. (2016). The transformation for GRI.MP9605 is only valid for g − i < 1.4. Sources redward of this limit were not used, and only stars (not galaxies) were used. The transformed Pan-STARRS stellar photometry was used to calibrate each CCD of each image separately. Occasionally, the number of stars on a CCD was insufficient to robustly calibrate that CCD. In these cases, the zero point was determined for the focal plane as a whole, and that zero point was applied to the individual CCD. Magnitudes were measured through two circular apertures whose size depends on the seeing; the diameters are 2× and 5× the full width at half maximum (FWHM) of the point-spread function (PSF). An average correction of the small aperture to the large aperture is computed for bright stars, and this correction is then applied to the smaller aperture magnitudes.

Stellar catalogs from overlapping images were used to check the photometry. The photometry of two overlapping images was consistent to 0.005 mag for the two r bands, comparable to the quoted accuracy of the PS-DR1 photometry. However, the photometry of the GRI.MP9605 band was found to be good to only 0.02 mag. This is due to two factors. First, the transformation between the Pan-STARRS g, r, and i filters and our wide filter is complex, with nonnegligible scatter. Second, the wide filter causes the PSF to vary depending on the spectral energy distribution of the stars: bluer stars will have larger PSFs, since seeing is worse at blue wavelengths, and DCR (Section 2.8.2) will stretch some stars away from the horizon.

We first consider the total area of sky that OSSOS surveyed. The on-sky footprint of the eight blocks described in Sections 2.1 through 2.6 (Figure 2) is a total of 169.62 deg². The total searched area for OSSOS is 155.3 deg², with a geometry that is summarized in Table 1. This slightly smaller region accounts for both the true pixel area of the MegaCam field of view and the few incompletely searched chips. Active pixels fill 91.64% of the region inside the outer boundary of the MegaCam mosaic: each of the 36 or 40 chips is 2048 × 4612 pixels, and with the 018689 pixel^–1 plate scale, the field of view is 0.9164 or 1.018 deg². Rare pipeline failures occurred in 1%–2% of cases. They were caused by bright stars or large galaxies that elevated the background light levels and reduced point-source detections to a level that precluded our automated PSF construction process. This rate is consistent with the sky coverage of the background contaminants. The effects are quantified by the "filling factor" in Table 2.

The TNO discovery pipeline and analysis techniques for OSSOS are exhaustively described in Sections 4 and 5 of Bannister et al. (2016b). The depth of each discovery triplet of images was quantified by planting PSF-matched sources in copies of each image at magnitudes m_r = 21.0–25.5. We planted 44k sources in each of 13AE and 13AO (Bannister et al. 2016b), of order 2 million sources in 13BL and 14BH, 700k sources in 15AP, and 400k sources in 15AM and all 15B blocks. We found empirically that about 10k sources per CCD were required to (a) provide the resolution needed to precisely determine the efficiency as a function of motion rate and (b) accurately measure the pipeline's ability to correctly determine the true flux of a source as a function of magnitude. (Sources were planted and candidates recovered in iterative loops to avoid planting saturation.) This substantial increase in the number of artificial sources meant that we could not vet 100% of all artificial candidates (as was done in Bannister et al. 2016b). Instead, we audited a subset of about 1000 artificial candidates per field (about 20,000 per block). For every frame examined, we performed an audit of the planted sources that we used in calibrating the detection system. These audited sources were treated identically to the detections from the actual data frames and used to verify our planting and search processes. To achieve a precise measurement of the structure of the detection threshold and the fill factor, we use the full sample of planted sources. We conducted tests to ensure that the vetting process did not introduce an additional bias by examining all planted sources for some blocks. These tests revealed that above about the 30% detection threshold (Petit et al. 2004), the vetting process does not remove valid detections. Our quantification of our false-positive rate is detailed in Bannister et al. (2016a, Section 5.1) and was 0 for 13AE and 13AO; in the later six blocks, we also have a negligible false-positive rate. Our false-negative rate is ∼0.75%. However, the process of vetting all artificial sources did introduce fatigue, which can affect the vetting. The increase in the number of planted sources and subsequent alteration of our review process was the only material change in our detection and tracking process.

OSSOS provides precisely quantified TNO detection efficiencies. The rate of recovery of sources planted in the images as a function of magnitude is modeled by the functional form given in Bannister et al. (2016b):

$$\begin{eqnarray}&&\eta (m)=\displaystyle \frac{{\eta }_{o}-c{(m-21)}^{2}}{1+\exp \left(\tfrac{m-{m}_{0}}{\sigma }\right)},\end{eqnarray} \tag{ 4 }$$

where η_o is the peak efficiency, c is the strength of the quadratic drop, σ is the width of the transition from a quadratic to an exponential form, and m₀ is the approximate magnitude of the transition. The peak efficiency η_o is roughly the efficiency at m = 21, in the case where $\exp ((21-{m}_{0})/\sigma )\ll \,1$. A few of the efficiency functions could not be fit correctly with the functional form of Equation (4), as the simplex minimization did not converge with the square function. They were instead fit with a double hyperbolic tangent,

$$\begin{eqnarray}&&\eta (m)=\displaystyle \frac{{\eta }_{o}}{4}\left[1-\tanh \left(\displaystyle \frac{m-{m}_{0}}{\sigma }\right)\right]\left[1-\tanh \left(\displaystyle \frac{m-{m}_{0}}{c}\right)\right],\end{eqnarray} \tag{ 5 }$$

as described in Jones et al. (2006) and are marked as such in Table 2. The efficiency function is determined using the intrinsic flux from the source, i.e., the planted source magnitude. The variation in the measured flux from the intrinsic value (caused by noise in the measurement process) is modeled as a function of source flux for each block. This variation is accounted for during our survey simulation process (see Bannister et al. 2016b, Appendix A). For completeness, we list m_50%, the r-band magnitude at which the efficiency function drops to 50%, which was often used in earlier surveys. However, the key limiting magnitude used throughout OSSOS is instead m_limit, the slightly deeper, but no fainter than m_40% (Petit et al. 2004), characterization limit: per Bannister et al. (2016b), "the magnitude above which we have both high confidence in our evaluation of the detection efficiency, and find and track all brighter objects." The efficiency function parameters are given for all of OSSOS in Table 2 and are distributed in machine-readable form in our Survey Simulator (see Section 4).

Detectability and the effective area of survey sky coverage are a function of the distant minor planets' sky motion rate. For each block, we report multiple efficiency curves in Table 2, each for a range of rate of motion. These different efficiency curves account for the changes in effective area, as faster-moving objects are more likely to shear off a given chip between the exposures of the discovery triplet. This produces a variation in area coverage as a function of rate, which is only ∼2% less for motion rates >10'' hr⁻¹. The flux limit for each rate is also different: trailing of fast-moving objects makes the flux sensitivity poorer, as the source is spread over more pixels in each exposure, pushing faint sources into the noise. We set an upper limit of 15'' hr⁻¹, as the survey's focus is on discoveries at heliocentric distances ≳10 au; at rates faster than this, the object would also feature significant trailing within an exposure, causing reduced detection efficiency (partly due to reduced signal-to-noise and partly due to our pipeline being optimized for roughly circular sources). The survey simulator (Section 4) accounts for this variation in sensitivity as a function of TNO sky motion rate.

2.9.1. TNO Tracking Efficiency

2.9.2. Sensitivity as a Function of Heliocentric Distance

2.9.3. Sensitivity to Retrograde Orbits and Interstellar Objects

The OSSOS primary discovery set is 840 TNOs, given in the catalog described in Table 3: characterized OSSOS minor planets. These minor planets were found at brightnesses and rates of sky motion that mean that they have fluxes brighter than our tracking limit, with well-quantified detection efficiencies. (This includes two close objects that were not tracked; see Section 2.9). The characterized discoveries are the set that should be used for modeling solar system structure. We summarize their orbital populations in Table 4.

For completeness, we list all 109 remaining detections in the catalog referenced in Table 3 as uncharacterized OSSOS detections. In this machine-readable catalog, there are three TNOs that have detection efficiencies that are quantifiable but differ from the efficiencies provided here for the main survey. They are designated with the prefix Col3N and were fast-moving discoveries in the incomplete first attempt at a discovery triplet on the 14BH block in 2013 (Section 2.2).

Table 3 (uncharacterized) also catalogs 106 TNOs from all blocks that are designated with the prefix u. These exceptionally faint objects have a detection efficiency that is poorly quantified, as their faintness at discovery placed them below their block's rate-of-motion–dependent characterization limit. The uncharacterized objects are severely afflicted by unquantifiable observational biases. The lack of quantified detection efficiency means that the only aspect of the tracked, uncharacterized TNOs that we can be sure of is their orbits. Of the uncharacterized TNOs, 21 could not be tracked beyond the ∼2 hr arc of their discovery triplet; these are designated with the suffix nt. It is possible that some of the untrackable nt sample may actually be false positives; see Section 5.1 of Bannister et al. (2016b) for an evaluation of the false-positive rate of our analysis. We recommend exceptional caution in any use of the Table 3 uncharacterized TNOs and strongly advise against simply combining them with the characterized discoveries when modeling solar system structure.

To fully exploit the discovery of a TNO, we must know both its discovery circumstances and its precise orbit. Extensive tracking is often required before such precise orbits can be determined; this is especially critical for resonant objects. The resonant orbits require extreme orbital precision (fractional semimajor axis uncertainty roughly <0.01%) in order to determine the resonance libration amplitude, which is a diagnostic of the resonance capture mechanism. An additional classification to those in Bannister et al. (2016b) is for Jupiter-coupled objects; these are defined by a combination of perihelion distance and Tisserand parameter,

$$\begin{eqnarray}&&{T}_{J}={a}_{J}/{a}_{\mathrm{TNO}}+2\sqrt{({a}_{J}/{a}_{\mathrm{TNO}})(1-{e}_{\mathrm{TNO}}^{2})}\cos ({i}_{\mathrm{TNO}}).\end{eqnarray} \tag{ 6 }$$

We do not attempt orbital classifications for the nt discoveries or objects with an arc only within their discovery lunation.

As described in Bannister et al. (2016b), we classify the TNOs into different dynamical populations (Gladman et al. 2008) based on a 10 Myr integration of clones of the best-fit orbit and two extremal orbits, which are determined from the available astrometry using the orbit-fitting method of Bernstein & Khushalani (2000). When all three clones show consistent dynamical behavior, we declare the object's classification to be secure. Precisely determining how long an observing arc, or the frequency of observations, is required to secure an orbit is not strictly possible because it depends on the TNO's dynamical behavior. The trans-Neptunian region is riddled with chaos and complex boundaries in orbital phase space. Orbits near the boundaries of resonances must be very precisely determined before the dynamical evolution can be accurately presented. In contrast, secure classifications for orbits farther from the resonance boundaries can be achieved with less precision. We do not know in advance where a given object's orbit is relative to these boundaries.

The high-precision and well-sampled tracking of OSSOS has allowed 734 of the 840 characterized orbits to be securely classified with the arcs we present here. This is summarized by population in Table 4. Additional arc was required to further refine the orbits for some insecurely classified objects. The orbits of the o3e and o3o TNOs that were initially reported in Bannister et al. (2016b) now have an additional two oppositions of more sparsely sampled OSSOS arc and accordingly updated classifications. Once an object was securely classified, we discontinued target tracking (occasionally, further serendipitous observations may have occurred, which we measure and include; see Section 2).

At the completion of the 2013–2017 observations, 103 characterized discoveries with well-observed arcs remain insecurely classified. The chaotic nature of the resonance boundaries means these objects' dynamical classifications might never become secure, because they transition between resonant and nonresonant behavior during the 10 Myr of the classification integration.

Fewer than 3% of the OSSOS discoveries were previously known, largely because of the greater depth of OSSOS compared to previous wide-field surveys. We designate 23 TNOs in Table 3 (characterized OSSOS) with the suffix PD, which indicates that during their tracking (Section 2), we made use of available pre-OSSOS astrometry to assist in securing their orbits. Pre-OSSOS astrometry was located in two ways. After the discoveries from each block were made, we checked against the predicted location of all MPC-listed TNOs that were within the sky coverage. We also tested for linkages to the short-arc discoveries by CFEPS (Petit et al. 2011).

There are OSSOS objects with pre-OSSOS data that we do not designate as PD. Several OSSOS discoveries were subsequently given MPC designations earlier than 2013–2015 on submission, due to linkages that the MPC was able to provide to unlisted or exceptionally sparse earlier astrometry, e.g., o3l28 (2006 QE₁₈₁). These do not have the suffix PD. Additionally, some objects were independently discovered during our survey. Object o3l39 (2016 BP₈₁) was found in Kepler space telescope observations of the K2 Campaign 8 superstamp field of Uranus (Molnár et al. 2018), and o3l18 (2010 RE₁₈₈) was reported by Pan-STARRS (Holman et al. 2015; Weryk et al. 2016). The orbits of these TNOs were already securely classified with our observed arcs by the time the other astrometry became available, and we do not designate them with PD.

The high cadence and astrometric precision of OSSOS maintains our sample uniformity across all our discoveries, including those designated PD. We tested for the effect of the prior astrometry on our classifications by integrating our arcs both with and without the earlier observations. Our orbital classifications are independent of the possibility that an object might be a PD: no classifications changed in class, though a few shifted between secure and insecure. Note that none of the PD objects are on high-perihelion, exceptionally large-a orbits (Section 5.4), where instead the observed length of arc would become the dominant source of orbital uncertainty. The PD TNOs are consistent in quality with the rest of the data set, rather than exhibiting greater orbital precision. This held true even in cases where the earlier astrometry extended the orbital arcs to a decade or more in length.

Our linkages to earlier discoveries show that for sparsely sampled arcs, even multi-opposition orbits in the MPC are sometimes misleading. An interesting case is o5p064PD (2001 FL₁₉₃). The 2001 discovery by the Deep Ecliptic Survey (Elliot et al. 2005) was with a pair of observations on March 27 and 28. Fourteen years later, 2001 FL₁₉₃ was recovered in three single observations in 2015 by the Dark Energy Survey (Gerdes et al. 2017). Having only six measured points defined 2001 FL₁₉₃'s orbit with high uncertainty, despite its 14 yr arc. The orbit was a rare low-inclination, a ∼ 50 au, outer-belt TNO beyond the 2:1 (see Table 6 of Bannister et al. 2016b). The OSSOS discovery o5p064PD links back to this previous data. However, we find instead that o5p064PD = 2001 FL₁₉₃ is correctly on an a = 43.1 au, e = 0.04, i = 18 orbit, which is that of a standard cold classical TNO.

A few MPC-listed objects were nominally within our coverage but fell in chip gaps during our discovery observations (e.g., 2000 FA₈ was in a chip gap during the 2015 April 12 discovery observations for the 15AP block). We recovered and reported astrometry where these TNOs were serendipitously included in our tracking observations.

The Centaur (a < 30 au) population and the q ≲ 30 portion of the scattering population undergo strong gravitational interactions with the giant planets, making them an unstable, transient population. The larger q portion of the scattering population also interacts with Neptune and undergoes significant orbital evolution as a result, but the more distant encounters allow for longer dynamical lifetimes. These unstable orbits are a mixture of a slowly decaying remnant of a vast population emplaced early in the solar system's history when the Oort cloud was built and a steady-state intermediary population leaking from a more stable Kuiper Belt or Oort cloud reservoir (Duncan & Levison 1997; Gomes et al. 2008; Dones et al. 2015). OSSOS provides 21 objects classed as Jupiter-coupled and Centaurs, and 38 scattering TNOs. Shankman et al. (2016) considered the size distribution of the scattering TNOs using 22 objects: 13 from CFEPS, two from A16, and the first seven OSSOS scattering discoveries. The available ensemble sample has now more than doubled. Lawler et al. (2018b) and Shankman (2017) found a break in the scattering size distribution and an intrinsic population of 9 × 10⁴ for H_r < 8.66 (D > 100 km), with the Centaur population two orders of magnitude smaller at ${111}_{-44}^{+59}$ (Lawler et al. 2018a).

The OSSOS survey has acute sensitivity to all objects exterior to the orbit of Saturn and thus provides a strong sampling of the inner-belt component of the TNO population in 37 au ≲ a ≲ 39 au. OSSOS detected 11 inner-belt TNOs above our characterization thresholds. Given our thorough ability to detect objects interior to 37 au and the complete absence of nonresonant/nontransient objects in the range 30 au < a < 37 au, we find that the present-day inner belt does not extend inward of a ∼ 37 au (Figure 5, top left). It is noteworthy that the strong lower perihelion bound of q ∼ 35 au is consistent across the entire Kuiper Belt. Only one TNO with a semimajor axis in the region 36 au < a < 48 au has a perihelion dipping just below, to q = 34.6 au (Figure 5, top left). The main structure visible within the inclination distribution in the inner belt is that of the ν₈ secular resonance, which destabilizes orbits in this semimajor axis range that have i ≲ 15° (Kavelaars et al. 2009). After accounting for this instability zone, the distribution of inner-belt inclinations continues to be consistent with being drawn from the same dynamically excited population as is present in the main belt. Further examination of the architecture of the inner belt and its relation to the main belt (J.-M. Petit et al. 2018, in preparation) is likely to provide significant insight into the processes that resulted in the currently observed structure (e.g., Nesvorný 2015b).

As of 2016 October, the MPC listed 484 TNOs with arcs where orbits as fit showed no resonance, with 42 au < a < 48 au and σ_a < 5%, from surveys other than those in the ensemble (Volk & Malhotra 2017). The ensemble presented here has 530 such main-belt objects, of which 421 are from OSSOS, and they have σ_a < 0.1%; the orbital precision is substantially better, allowing for much more secure dynamical classifications. We show these in Figure 5. In the main belt, even the first two blocks of OSSOS were able to independently confirm the existence of the cold classical "kernel" (Bannister et al. 2016b), which was first described in Petit et al. (2011). With the full ensemble sample, the kernel is starkly visible in Figure 5. The tight parameter space of the kernel constrains the processes that are viable candidates for its sculpting out of the primordial disk (Gomes et al. 2018) or its emplacement (Nesvorný 2015a).

5.2.1. The Cold Classical Population Extends Beyond the 2:1 Resonance with Neptune

5.2.2. The Detached Component

OSSOS provides a substantive additional sample of TNOs in mean-motion resonances with Neptune (Figure 4). The four new and secure Neptune Trojans from OSSOS exhibit the known wide range of inclinations of this population (Chiang et al. 2003; Sheppard & Trujillo 2006, 2010; Parker et al. 2013; Alexandersen et al. 2016), discussed further in H. W. Lin et al. (2018, in preparation). A spectacular 132 plutinos (3:2 resonance) dominate the resonant detections of OSSOS; combined with the survey simulator, the detailed debiased distribution will provide the information needed to study and constrain resonance capture conditions and plutino mobility over the age of the solar system (K. Volk et al. 2018, in preparation).

Observational biases make the TNOs in more distant resonances progressively harder to detect, as they spend a larger fraction of each orbit at greater heliocentric distances, where they are fainter. For example, OSSOS found only 34 TNOs in the 2:1 resonance. However, early OSSOS findings support earlier findings that the total population in the more distant resonances may rival or exceed those of the nonresonant Kuiper Belt (Gladman et al. 2012). Volk et al. (2016) found that the OSSOS o3e and o3o discoveries imply that the 5:2 resonance (a ∼ 55 au) is much more heavily populated than had previously been suggested from migration models. The significant number of objects OSSOS has observed in some of the resonances will enable detailed modeling of their dynamical structure and relative populations, providing detailed constraints on Neptune's migration history. For example, the o3e and o3o discoveries confirmed that the objects in the 2:1 have a colder inclination distribution (Gladman et al. 2012) than the population of objects in the 3:2 (Volk et al. 2016). This is now visually obvious in Figure 4 and is discussed further in Y. T. Chen et al. (2018, in preparation).

The number of distant resonant objects beyond the 2:1 listed in the MPC from other surveys is ∼103, contrasting with 68 in the ensemble, of which 48 are from OSSOS. Table 4 highlights that OSSOS provides the first TNOs found in a number of distant resonances. These include occupancy in the a ∼ 82 au 9:2 resonance and the most distant securely resonant TNOs yet seen, in the a ∼ 130 au 9:1 resonance (Volk et al. 2018). OSSOS has also found additional TNOs in other rarely seen distant resonances: seven in the 3:1 and three in the 4:1 resonances (Chiang et al. 2003; Alexandersen et al. 2016). The largest TNO found by OSSOS, the H_r = 3.6 ± 0.1 dwarf planet candidate o5s68 (2015 RR₂₄₅), is in the 9:2 resonance (Bannister et al. 2016a); its long-term behavior suggests that a large metastable population is cycling between the scattering population and high-order resonances.

The OSSOS discoveries include nine TNOs with q > 30 au and a > 150 au, eight of which have q > 38 au. Their orbits are shown in Figure 6. Pronounced biases affect the discovery of minor planets on such distant orbits; the sensitivity of OSSOS to this population is quantified in Shankman et al. (2017). There has been recent interest in the apparent angular clustering of the MPC-listed TNOs with a > 150 au orbits, which some have hypothesized as evidence for a massive distant planet (Trujillo & Sheppard 2014; Batygin & Brown 2016). The sensitivity of OSSOS was applied to a distribution of simulated orbits that are uniform in their angles ω, Ω, and ϖ to test if the detected OSSOS sample is consistent with being drawn from a uniform distribution. In an analysis of the orbit distribution of the first eight of these TNOs, Shankman et al. (2017) found that the OSSOS sample is consistent with being detected from that uniform distribution.

Zoom In Zoom Out Reset image size

We report a new minor planet in the q > 30 au, a > 250 au population. The TNO uo5m93, an object in the uncharacterized OSSOS sample (Table 3), was eventually tracked to a well-sampled four-opposition arc and found to have an orbit with q = 39.5 au and a = 283 au (Figure 6). Thorough inspection of the complete detection sample has minimized the possibility that any further q > 30 au, a > 150 au orbits remain concealed among the OSSOS nt discoveries: fewer than 16 (with d > 30 au) of the 111 uncharacterized objects were untrackable (see Table 3: uncharacterized, nt designations). The spatial orientation of uo5m93's orbit lies in an angular direction that is distinct from the hypothesized clustering, further weakening the evidence for a lurking presence of intrinsic angular clustering.

Although this object is from the uncharacterized part of the sample, where our completeness limits are less well known, we reperformed the full statistical analysis of Shankman et al. (2017); the results are unchanged, and the OSSOS q > 30 au, a > 150 au nine-TNO sample is still consistent with being detected from a distribution of orbits that are intrinsically uniform in the angular elements.

Formation mechanisms for this distant population are not yet clear, and it remains an area of active investigation (e.g., Lawler et al. 2016; Nesvorny et al. 2017). However, all of the extreme TNO discoveries of OSSOS are consistent with a formation by random diffusion in semimajor axis due to weak kicks at perihelion by Neptune from orbits with semimajor axes in the inner fringe of the Oort cloud, as proposed in Bannister et al. (2017). This is shown in Figure 7, which updates Figure 5 in Bannister et al. (2017): the OSSOS discoveries are in the phase spaces that can be populated by the combined mechanisms of diffusion, scattering, and capture into weak distant resonances. We refer the reader to Bannister et al. (2017) for detailed discussion.

Zoom In Zoom Out Reset image size