The DP2 window, fourteen nights from MJD 60865 to 60878, linked as overlapping cones. Each patch is searched twice: the main-belt grid, then the near-Earth grid on whatever the first pass did not claim. 247 of 254 patches finished. Cones above 4,950 detections per square degree, the bulge, are not in this run. The orbit of every pure linkage is drawn here.
The sky under the density cut is 2,668 cones of radius 2.5°, grouped into 254 patches. Neighboring cones share a job, and a patch that would hand the linker more than 180,000 detections is shrunk. The seven patches that did not finish are absent from every total below. Patches overlap, so a sum of their detection counts is not a count of the sky.
The first pass uses heliohypo_rubin324, heliocentric distances
from 1.5 to 9.8 AU. Every detection that landed in a refined linkage is then
removed, and the second pass uses heliohypo_neo324, from 0.85 to
1.5 AU, with the innermost geocentric bin at 0.01 AU. Both passes use the
same tracklet limits: up to 0.6° per day, pairs from 0.3 to 120 minutes,
a clustering radius of 50,000 km, and a 1″ residual cut. A state is
dropped only when both ends of its tracklet fall inside 0.05 AU and the
straight-line miss is under 20,000 km. That cut is there to discard the
hypotheses that pile up on top of the observer. It does not change the
main-belt pass, which never gets that close.
The planted objects are the brightness draw: each one was placed bright enough to be detected somewhere in the window. A draw by diameter would put many more faint objects into the denominator. The catalogue is opened only to score the linkages.
A finished patch took a median of 2.9 minutes. 236 of the 247 finished in under ten minutes, and the slowest finished one took 85 minutes. The main-belt linker had a median of 42 seconds. The near-Earth linker had a median of 30 seconds and a maximum of 79 minutes, so the second pass is cheap on a typical patch and expensive on a few. The 247 finished patches are 135 core-hours. The seven that hit the two-hour cap are another 112.
The finished patches contain 238,069 catalogued objects with at least one detection. 69,861 were recovered, 69,704 of them on the main-belt pass. The near-Earth pass added 157 that the first pass had not recovered.
The denominator used above is narrower. An object counts as eligible when its own tracklets cover three or more nights. That is 83,549 catalogued objects, and 69,585 of them were recovered. The other 13,964 eligible objects are the linker's miss. Most of the objects that were not recovered never reached that denominator:
| Stage | Objects |
|---|---|
| Outside tracklet limits | 96,442 |
| Mispaired | 57,697 |
| Never clustered | 11,605 |
| Two nights clustered | 2,197 |
| Dropped by purify | 247 |
| Crowded cluster | 20 |
Outside the tracklet limits means fewer than three clean nights and no impure tracklet. Mispaired means an impure tracklet and still fewer than three clean nights. Those two rows are cadence, pairing, and the rate window. The remaining rows are objects the linker had in hand and did not return as a pure three-night cluster.
251,916 orbits were planted, the full brightness draw. 201,661 of them produced a detection that entered the catalog, beside 238,069 catalogued objects in the finished patches, 85% of that population. A class-by-class cut aimed at 5% of the real detections was built for the middle-third strip. It was not applied to this run.
The outcome is the first that applies, scored on the whole window. A detection outside the planting field counts. Below depth, lost to trailing, and depth unknown are orbits that came near a visit and never cleared it. No tracklet means the orbit was detected and still has fewer than three nights with a pair. Lost to the reliability cut means those three nights existed before the cut and not after.
| Outcome | Orbits |
|---|---|
| Never observable | 390 |
| Below depth | 48,570 |
| Lost to trailing | 241 |
| Depth unknown | 10 |
| No tracklet | 64,137 |
| Lost to the reliability cut | 4,926 |
| No linkage | 45,941 |
| Recovered | 87,701 |
The linker recovered 87,717. 16 of those had fewer than three paired nights in the catalog, so the outcome table files them under an earlier row and counts 87,701 as recovered.
201,620 planted objects have a detection in a finished patch, and 87,717 were recovered. 86,917 of the recoveries are the main-belt pass. The near-Earth pass added 800. Eligible recall, three tracklet-nights, is 87,690 of 102,483. The stage table under that is every such object the search did not recover, including orbits that never had three paired nights. No linkage, in the outcome table, is the narrower set: three paired nights, and still no recovery.
113,903 were seen and not recovered. The stage is the furthest any finished patch reached, and the names are the ones in the catalogued table:
| Stage | Objects |
|---|---|
| Outside tracklet limits | 58,785 |
| Mispaired | 40,250 |
| Never clustered | 13,386 |
| Two nights clustered | 1,145 |
| Dropped by purify | 325 |
| Crowded cluster | 12 |
By class, among the eligible objects:
| Class | Eligible | Recovered | Recall | Near-Earth pass only |
|---|---|---|---|---|
| Apollo | 48,256 | 42,864 | 0.888 | 119 |
| Amor | 47,730 | 41,700 | 0.874 | 594 |
| Aten | 6,497 | 3,126 | 0.481 | 78 |
| All | 102,483 | 87,690 | 0.856 | 800 |
Apollos and Amors that form three nights of tracklets are recovered at about the same rate as the catalogued objects. Atens are not. 33,123 Atens were seen and only 6,497 of them formed three tracklet-nights, so the loss is mostly before linking. Of the Atens that did, about half came back, and 78 of those 3,126 recoveries are the near-Earth pass. The second grid is a small addition on top of the main-belt pass, not a second harvest. It is where the close objects the first grid cannot hold have to be found, and on this window that was 800 planted objects and 157 catalogued ones.
A mixed linkage contains more than one catalogued or planted object. An unlabelled one contains neither. Patches overlap, so the same cluster is written several times. The counts here keep one copy of each set of detections.
| Pass | Objects in a pure cluster | Mixed clusters | Of which hold a planted detection | Unlabelled clusters |
|---|---|---|---|---|
| Main belt | 156,643 | 1,003 | 575 | 14,941 |
| Near-Earth | 2,124 | 365 | 283 | 136 |
On the main-belt pass the mixed clusters are rare next to the pure recoveries, about one mixed cluster for every 156 objects in a pure cluster. On the near-Earth pass, 365 mixed clusters sit beside 2,124 objects in a pure cluster. The candidate list from the second pass is dirtier. A recovery itself is still a pure cluster: the mixed rows are other linkages, not the objects counted as found.
Four patches in the south have a single night of detections, so a three-night linkage was never available. The main-belt pass returned nothing, which is the right answer, and the near-Earth pass was still running at two hours. They are patches 10, 14, 15, and 17, between declination −40.5° and −45.5°, with 92,000 to 175,000 detections.
Three patches near the ecliptic finished the main-belt pass and died in the near-Earth pass. Their main-belt linkages are on disk and are not in the totals above: patch 139 has 2,008 objects in a pure cluster, patch 155 has 605, and patch 156 has 1,014. Patch 155 got through the near-Earth linker in 93 minutes, wrote 66,575 clusters, and ran out of time while that file was being written. Patches 139 and 156 were still inside the linker, at 7.2 GB and 8.6 GB. The same two-hour cap is what stopped them.
Source: output/linking/cross_cone_neo_full on the
cluster, 247 of 254 patches, merged 10 October 2026. Object
counts take the best stage any finished patch reached. Cluster counts
keep one copy of each detection set.