NASA Lost This Asteroid — Then It Came Back as Something Else

[00:00:00]On August the 26th, 2025, Goldstone aimed its 70 m radar dish at 1998 SH2.

[00:00:09]The NASA and JPL beam came back empty.

[00:00:13]The orbit placed the object inside a 70 arcsecond radar beam [music] and an echo was expected within minutes that night.

[00:00:21]None appeared in 40 minutes. 5 days later, S O N E A R found 1998 [music] SH2 153 arcsec from the gravity only prediction outside the beam. It had [music] not vanished. Something in the model still placed it in the wrong place. Goldstone had reasons to trust its equipment. A transmitter problem had limited the dish [music] to 240 kW, but the radar had detected another asteroid before [music] the attempt. About 20 minutes into the 1998 SH2 track, the team [music] checked the data at several frequency resolutions. Still nothing.

[00:01:05]After roughly 40 minutes, they ended the [music] attempt and moved to another asteroid. That target appeared. Radar [music] had not failed completely. It had spent its narrow window interrogating a piece of empty sky. This is not a story about an asteroid disappearing. [music] It is a test of instrument pointing orbit and physical identity. The answer would require a recovered point of light, [music] a second radar track, and images deep enough to reveal what routine survey pictures [music] could not. The record begins on September 17th, 1998 when Spacewatch first observed the object later numbered 8 7 5163 [music] and known as SH2.

[00:01:53]That line began the ark. An orbit is built from positions measured against the sky. One observation gives a point.

[00:02:01]Repeated observations show a path. From 1998 through October 2016, astronomers collected [music] 148 optical measurements of 1998 SH2.

[00:02:15]Fit those points under the pole of the sun and planets and the path [music] can be projected forward. The result is not a painted line. It is a corridor of probability that widens with time, measurement quality, and the geometry of the observations.

[00:02:32]On paper, SH2 looked like an asteroid, a dark Apollo class near Earth object a few hundred meters across. Neoise data suggested a diameter near 380 m and a very low albido, meaning its surface reflected little sunlight. [music] Its orbit also carried a clue, a quantity called the Tiseran parameter placed it near the dynamical territory of Jupiter family comets. But dark [music] surfaces and comet-like orbits are not verdicts.

[00:03:04]An extinct [music] looking comet nucleus and an ordinary dark asteroid can resemble the same unresolved [music] point. Those properties kept a cometary ancestry possible, but neither measurement could show active dust or gas around the body. That distinction matters because a radar beam is not [music] the whole sky.

[00:03:25]Goldstone's DSS14 antenna is 70 m across. For these observations, it transmitted at a wavelength of about 3 1/2 cm. The energy was concentrated into a primary beam only 70 arcsec wide. That is the strength of planetary radar.

[00:03:44][music] Place a body inside the beam and the returning echo can reveal distance, motion, and sometimes shape. The same concentration is also the weakness. Put the target outside the primary beam and an enormous dish may return [music] nothing from the coordinates it interrogates.

[00:04:04]The gravity only solution gave [music] a formal three sigma pointing uncertainty of plus or minus 24 arcseconds. [music] That region sat comfortably inside the 70 arcsec beam. An echo should have appeared within minutes. It did not. Yet the radar saw neighboring targets. A failed instrument demands repair.

[00:04:25][music] A failed prediction demands that the coordinates be earned again. A blank result is only useless if the [music] question that produced it was right.

[00:04:36]Here, the blank screen said the team needed another question and another observation before it could trust the old coordinates again.

[00:04:45]Time had been working against that [music] confidence. No new telescope observations entered the study's tracking arc after October 2016. During the gap, 1998, [music] SH2 completed two circuits around the sun. If the model [music] lacked even a very small persistent force, the error would not announce itself with [music] a dramatic turn. It would accumulate. Each orbit would carry the predicted point a little farther from the real one, while the old uncertainty calculation remained precise inside the assumptions it had been given. The absence [music] of new points did not make the orbit wrong. It made any missing physics harder to notice.

[00:05:26]The opportunity in August 2025 [music] was valuable because 1998 SH2 passed relatively near Earth.

[00:05:36]Current JPL data [music] placed closest approach on August 30th at about 3.11 million km, roughly eight times [music] the Earth moon distance. It was a safe pass. The object meets the technical category of a potentially hazardous asteroid [music] because of its size and orbital proximity, but that category is not a collision forecast. The value of the encounter [music] was measurement, not alarm. If you think details like this deserve a closer look, subscribe. We follow the evidence wherever it leads.

[00:06:10]Goldstone had expected a moderate radar echo during [music] that close approach.

[00:06:15]Instead, the formal uncertainty sat inside the beam while the real object sat [music] outside it. Repeating the same coordinates on September 2nd would risk repeating the same failure. Before radar could measure the body, an optical telescope had to tell the dish where to look. The normal [music] picture had failed its first physical test. On August 31st, observers at the S ne facility in [music] Brazil recovered a point of light. It was the first tracking of SH2 since 2016. The measured position was not a small adjustment at the edge of the old uncertainty. It was 153 arcseconds from the gravity only prediction, about 2 1/2 ark minutes [music] and 19 times the quoted one sigma scale in the study's analysis.

[00:07:05]That new position arrived just 2 days before Goldstone's next scheduled [music] attempt.

[00:07:10]Without it, the second radar window could have opened on the same empty [music] coordinates. The new point did more than recover a name. It reopened a measurement opportunity. Put the three numbers on one diagram. The formal three sigma uncertainty reached 24 arcseconds from the predicted [music] center.

[00:07:32]Goldstone's main beam was 70 arcsec [music] wide. The recovery point lay 153 arcsec from the prediction. Even without deciding what physical force caused the difference, the geometry closes the first question. The object had been outside the interrogated patch of sky.

[00:07:52]This was not a subtle miss at the beam's edge. It was a target beyond the main lobe. Reduced transmitter power may have weakened the expected echo, but it could not move the primary beam onto a target recovered 153 arxicons from the predicted center. One limitation affected signal strength. [music] The other placed the strongest part of the beam on the wrong sky. On September 2nd, [music] DSS14 tried again with the updated ephemeris.

[00:08:22]This time, the echo appeared in less [music] than two minutes. The team began with a continuous wave observation, [music] then moved to coded waveforms that could measure range. The full radar sequence lasted about 67 minutes.

[00:08:35]Nothing in that echo showed a tail or proved a comet. Radar [music] solved the operational problem. Revised coordinates put a real target back into the beam.

[00:08:45]The successful return also paid a debt left by August 26th. The antenna had not watched an object vanish. It had been pointed to the wrong place. Set August 26th [music] and September 2nd side by side. The antenna and the object were the same. The pointing solution was not.

[00:09:05]The first date produced [music] roughly 40 minutes without the target. The second produced an echo in under 2 minutes and then a 67-minute observing sequence. That contrast does not identify a comet or measure out gassing.

[00:09:19][music] It establishes something more basic. A fresh sky position restored the measurement. The transmitter limitation remains part of the first night record without becoming the whole explanation.

[00:09:32]Lower power affected the expected echo.

[00:09:35]Updated coordinates changed where the primary beam went.

[00:09:39]One corrected track could [music] still leave room for a bad optical measurement or a short-lived error. But in Faroia and colleagues analysis, [music] the recovery did not stand alone. They added more than 200 optical observations and the radar delay measurement to their orbital fit. In that [music] fit, a non-gravitational term remained. Better data made a single bad position harder [music] to sustain as the explanation within their model. The investigation was no longer only about one telescope finding one faint point [music] beyond a gravity only prediction. It was also about whether the study's added force term represented real physics and what observation could test [music] that interpretation.

[00:10:20]Then the optical evidence seemed to resist the most obvious new identity.

[00:10:26]Atlas collected 53 [music] images from September 3rd into early October.

[00:10:31]Individual exposures showed no coma and no [music] tail. Even a stack from September 3rd and 4th still looked point-like. That negative result deserves its full weight. If outgassing had pushed the body, where was the dust?

[00:10:47]But an image can only rule out what [music] it was sensitive enough to show.

[00:10:52]A point source in a survey [music] frame is not proof that no fainter structure exists beneath the surface brightness limit. The first plausible explanation preserved [music] the asteroid picture.

[00:11:04]JPL's living goldstone [music] planning page identified a transverse acceleration term A2 and initially described it as [music] the Yarovsky effect. That was reasonable. A dark rotating body can absorb [music] sunlight, warm, and return that energy to space unevenly.

[00:11:24]The escaping infrared photons carry momentum. Over years, their tiny recoil can alter an asteroid's orbit. The sign of the fitted term even suggested probable retrograde rotation. The [music] question was not whether Yarovsky was real. It was whether it was large enough here. Picture one side of the body heating in [music] sunlight.

[00:11:46]Rock and dust do not release all that heat at local noon. The warmest region comes later after rotation has carried the surface away from the sun. That delay gives the departing infrared energy a preferred direction and the recoil [music] can gently speed or slow the body along its orbit. No engine is involved. No visible plume is required.

[00:12:09]Across years and millions of kilome, a [music] force too small to feel can still become an astrometric correction worth measuring. Now compare the scale [music] inside that paper's analysis.

[00:12:22]Farakia and colleagues fit returned a transverse term equivalent to about - 1.4 [music] * 10 to the -1 m/s squared.

[00:12:32]Their maximum modeled Yarovski acceleration [music] for a body like 1998 SH2 was about 10 times smaller. These were not two direct force measurements. [music] Both were quantities computed within the study's assumptions. Yet within that analysis, the comparison was not close.

[00:12:51]thermal recoil [music] could contribute, it could not supply the full fitted term. A familiar asteroid mechanism survived as real physics [music] and failed as a complete explanation of the author's result. That numerical mismatch changed the weight of the evidence inside the study. The object had missed the old prediction [music] by 153 arcsec. A revised ephemeris had returned an echo in under 2 minutes. more than 200 later optical observations and the radar delay preserved the added term in [music] the author's fit. Ordinary survey images still showed a point, but their physical explanation now had to do more than keep the asteroid label comfortable. It had to account for the scale of their fitted term [music] and survive a test outside the orbit calculation. The orbital fit itself [music] did not identify a gas. It solved for a generic transverse term called A2, a compact way to represent a small acceleration along the orbit. After finding Yarovsky too weak in their modeling, Farinokia and colleagues favored weak outgassing [music] as the physical source. Material leaving a surface would carry momentum in the opposite direction, like a nearly invisible thruster whose strength changes with solar heating. The later activity images [music] make that interpretation physically plausible.

[00:14:14]They do not turn it into a direct measurement of historical thrust. No gas species was measured. No buried ice reservoir was seen. And [music] the FIT could not uniquely replay every moment between 2016 and 2025.

[00:14:30]That left a harder test. If the outgassing [music] interpretation was physically right, the body might have lifted dust too faint [music] and diffuse for Atlas to resolve. The Danish 1.54 m telescope [music] collected images in midepptember.

[00:14:46]The Canada France Hawaii telescope observed on September 17th and 18th. The Very Large Telescope followed on September 30th. [music] Their larger apertures, longer integrations, and careful stacks could combine light spread thinly [music] across many pixels. The question had moved beyond a term in an orbit fit. Did the [music] body itself leave a visible commentary structure? In the deep stacks, it [snorts] did. The CFHT, [music] VLT, and Danish telescope images resolved a faint coma [music] extending at least 10 arcsec around 1998.

[00:15:25]SH2 [music] and a narrow tail reaching at least 20 arcseconds. These were not features in the Goldstone Echo [music] and they were not visible in the routine atlas frames.

[00:15:36]They emerged when the observations reached [music] lower surface brightness.

[00:15:41]That is the direct evidence. Whatever uncertainty remains in the orbital force history, a resolved coma and tail establish that the object was active.

[00:15:52]The asteroidon physical label no longer described [music] everything in the image. Place the early atlas stack beside the later deep images. In the first 1998 SH2 remains a point. In the second, [music] light extends around it and trails away. The contrast does not [music] establish that activity suddenly began between those dates. It shows that deeper observations could separate faint [music] diffuse light from the sky background where routine survey frames [music] could not. That is why 53 blank survey images and the later tail can both be true without weakening either observation. The tail can be modeled too, [music] but the evidentiary line changes again. Its direction and grain distribution were consistent [music] with dust released from late August into early September rather than in one instantaneous [music] collision. The authors favored sustained sublimation and discussed delayed, probably subsurface activity. Yet, the images do not reveal a named volatile or expose an ice layer. They do not mark the exact first or final moment of [music] a mission. The tail proves activity. Its detailed history remains [music] a reconstruction from dust, sunlight, and observation geometry. So, the evidence now has a visible edge established. The old beam was empty. The body was recovered 153 arcsec.

[00:17:29]The revised ephemeris returned it to radar [music] and deep images showed a coma and tail. In Farochia and colleagues analysis, an A2 term was fitted and their maximum modeled Yarovsky effect was about 10 times too small. [music] They favored weak outgassing as the physical explanation.

[00:17:47]Still unmeasured, the volatile, the subsurface structure, the complete activity window, and whether any comparable acceleration repeats on another return. The object's commentary [music] nature is more secure than any single reconstruction of the force history that carried its prediction across the sky. Now return to August 26th. The blank record means something different after the tail image and the [music] force comparison.

[00:18:14]Goldstone had not shown where the object was. It had shown where confidence in the old prediction ended. The dish accurately examined coordinates produced [music] by a gravity only solution.

[00:18:26]reality lay 153 arcsec away. Whatever a future orbit fit concludes about the exact [music] force history, the cost of that incomplete prediction was not abstract. 40 minutes of radar time [music] spent on empty sky, then an optical recovery and a new orbit before the second window could be used. That does not make SH2 an incoming danger. It remains in the potentially hazardous category, but the paper reports zero foreseeable Earth impact probability in its [music] analysis. The lesson is about better prediction, not panic. For near-earth objects, position uncertainty depends on the length and quality of the observation arc, its geometry, and the forces a chosen solution tests. A [music] weekly active comet that looks point-like in routine images can matter because its physical nature tells modelers which [music] effects deserve examination.

[00:19:26]By July 2026, the documentary record used several labels at once. [snorts] The Nature Astronomy paper described the object as dual status [music] and used SH2.

[00:19:40]NASA JPL's July 16th account [music] said it would receive that provisional comet designation.

[00:19:47]JPL's radar history archive [music] marked the target comet while the minor planet center had issued a circular titled [music] commentary activity N875163.

[00:19:59]Those sources use different administrative wording. None of it changes the [music] physical evidence.

[00:20:05]The decisive record is not the label. It is the resolved coma and tail. What remains needs another return, not a louder claim. Astronomers have not directly measured which volatile drives the activity. They do not yet know the full duration, whether the behavior repeats each orbit, [music] or which time dependent force law will best predict the body over future years. Nor does one reclassified object tell us how common such cases are. Those answers require observations at different points in the orbit, including periods when a routine survey may again see nothing [music] but a dot. Future observations must test those blank spaces directly.

[00:20:50]The important open question is therefore larger than the name attached to 1998 SH2. [music] How many near-Earth bodies cataloged from pointlike images have [music] activity too faint for ordinary survey frames yet strong enough to alter a future position? At present, [music] there is no defensible count. Farncia and colleagues [music] present this as the first case in which a fitted non-gravitational perturbation [music] predicted cometary activity and targeted observations then confirmed it. One case is not a hidden population. It is a method. Let an unexpected trajectory identify which quietl looking objects [music] deserve a deeper image. The method scales without pretending that the first example supplies the answer.

[00:21:41]Return once to the first diagram.

[00:21:42][music] Draw Goldstone's beam as a circle 70 arcsec [music] wide. Then place the recovered point 153 arcseconds from the old prediction. [music] On August the 26th, the circle contained no [music] echo. 5 days later, the point appeared beyond its edge. The radar had worked well enough to detect neighboring targets, and the object had remained real. No adjustment after the fact can pull that measured point back inside the circle that missed it. The dish returned nothing from the place Confidence had drawn. Reality continued outside the circle.