Planet Nine Searches

Speculative If the orbital clustering in the extreme trans-Neptunian objects is a real feature of the sky rather than a map of where telescopes happened to point, the prize is a planet of about 6.2 Earth masses at roughly 380 astronomical units — the first planet added to the Solar System since Neptune in 1846, and the nearest example of the commonest planet class in the galaxy. The condition is doing real work in that sentence: the same data yield a 0.2% chance of coincidence under one debiasing method and a 17% to 94% agreement with pure randomness under another, and 3 is the longest part of this brief because that disagreement, not the planet, is the actual state of the subject.

Established The unconditional prize is already being delivered, and it is a survey rather than a planet. The whole astronomical community catalogued roughly 5,000 trans-Neptunian objects over about thirty years; the Vera C. Rubin Observatory added 380 of them in approximately six weeks of commissioning, and its ten-year Legacy Survey of Space and Time is predicted to find about 37,000 — a factor of seven on the existing census, with roughly 70% of the distant populations arriving in the first two years. All of them come from one telescope on one cadence, so for the first time the detection probability of a distant object can be computed rather than assumed.

Frontier The gain is public and the loss is cheap, which is unusual for a question this loud. The clustering argument currently rests on 14 objects, and the characterised subsample that both camps agree is analysable is smaller still; a fourfold increase in that number is enough, by the field’s own simulation work, to move the question from arguable to decided. No new instrument and no new launch are required: the survey that settles it was funded to measure dark energy and supernovae, and the outer Solar System result is a by-product of a cadence chosen for other reasons.

Speculative The condition, stated plainly: the upside survives only if a single characterised selection function can be applied to the whole sample, and today no such thing exists. The 14 objects that carry the clustering claim were found by surveys with different depths, different sky coverage and, in at least one case, a pointing strategy chosen because of the hypothesis under test. 5 is about that problem and nothing else. If the clustering dissolves, what is left is still worth the money — a calibrated map of everything between 30 and 2,500 astronomical units — but it is a smaller and quieter prize than the one the field has been advertising for a decade.

Established The subject is a claim about angles, not about a photograph. Beyond Neptune, at 30 astronomical units, lies a disc of icy bodies; a small subset of them have semimajor axes of hundreds of astronomical units and perihelia well outside Neptune’s reach, so nothing known can be keeping them on those orbits. These are the extreme trans-Neptunian objects, or eTNOs. Their individual orbits are measured superbly well. 2017 OF201, announced in 2025, has a semimajor axis of 830.1 ± 0.8 au, a perihelion of 44.9189 ± 0.0012 au and an eccentricity of 0.94589 ± 0.00005. The dispute is not about any one orbit. It is about whether the directions those orbits point in are distributed evenly around the Sun.

Frontier Three angles carry the argument, and they are the three that a flux-limited survey is worst at sampling fairly. The argument of perihelion ω says where in its own orbital plane a body comes closest to the Sun; the longitude of ascending node Ω says where that plane cuts the ecliptic; the longitude of perihelion, their sum, says which way the long axis of the ellipse physically points in space. Trujillo and Sheppard reported in 2014 that distant objects cluster in ω near zero. Batygin and Brown argued in 2016 that the physical direction of the orbits clusters too, and that a distant planet anti-aligned with them would hold them there.

Established Every one of those objects was found near its own perihelion, because that is the only time it is bright enough to see. An object on a 830-au orbit spends almost all of its period beyond a thousand astronomical units and is detectable only in the small fraction of it spent near the Sun.

Established A survey therefore samples a distant population not by orbit but by where the survey was looking and how deep it went, and the observed angular distribution is the intrinsic one multiplied by a selection function. Recovering the first from the second is the entire technical content of this subject.

Speculative “Planet Nine” names a specific dynamical solution to that pattern, and the pattern has at least four other candidate explanations. A planet of 5 to 10 Earth masses on a distant eccentric inclined orbit is one. A self-gravitating disc of small bodies massive enough to shepherd itself is another. A collective instability in a primordial scattered disc is a third. A modification of the law of gravity in the weak-field limit is a fourth. A primordial black hole of the same mass is a fifth, and differs from the first only in radius. 4 takes each at its honest level. All five are downstream of the same prior question: whether there is a pattern at all.

Established Start with what nobody disputes: the orbits. Astrometry of distant minor planets is one of the most precise measurements in observational astronomy, and the numbers are quoted to more digits than anything else in this brief. 2015 BP519, found in Dark Energy Survey data, has a semimajor axis of 448.99 ± 0.49 au, an eccentricity of 0.92149 ± 0.00009, a perihelion of 35.249 ± 0.078 au and an inclination of 54.1107 degrees. 2017 OF201, recovered from Dark Energy Camera Legacy Survey archives across detections spanning 2004 to 2025, has an aphelion of 1615.2 ± 1.5 au and an estimated diameter of about 700 km at an assumed albedo of 0.13. These objects exist, their orbits are secure, and no part of the controversy touches them.

Frontier The clustering claim has been restated four times with four different numbers, and the numbers have moved in both directions. Batygin and Brown’s 2016 paper took six objects with semimajor axis greater than 250 au and perihelion beyond 30 au, and reported that “orbits as tightly clustered in perihelion position as the 6 observed KBOs occur only 0.7% of the time”, with a joint probability against the pole orientations of 0.007% — about 3.8 sigma. Brown’s 2017 follow-up debiased ten objects beyond 230 au against the discovery circumstances of 1,248 catalogued Kuiper belt objects and got 1.2%.

Frontier Brown and Batygin in 2019 extended that to the 14 objects known on 1 July 2018, reporting 4% for the longitude of perihelion alone, 3.5% for the orbital poles alone, and 0.2% jointly. Their 2021 orbit paper quotes 99.6%.

Frontier The commission for this brief cited “~99.8% significance (Batygin & Brown)”, and that figure is right but its attribution is off by a paper and three years. The 0.2% joint probability — 99.8% confidence — is from Brown and Batygin’s 2019 Orbital Clustering in the Distant Solar System, not from the 2016 discovery paper, which reported the much stronger and much smaller sample figure of 0.007%. The distinction matters: the headline significance has weakened as objects have been added, which is the opposite of what a real signal usually does, and the 2021 restatement at 99.6% weakened it again. Anyone quoting a single number for this claim is quoting a snapshot of a moving quantity.

Established The rebuttals are not opinions; they are quantified results from surveys that published their own selection functions. The Outer Solar System Origins Survey observed 155 square degrees with the Canada-France-Hawaii Telescope between 2013 and 2017 to depths of m_r = 24.1 to 25.2, injected millions of synthetic sources to measure its own detection efficiency, and tracked 838 of 840 discoveries to orbits that “can never be lost”.

Established Shankman and colleagues then applied that selection function to the eight OSSOS objects with a > 150 au and q > 30 au, four of them beyond 250 au, and ran Kuiper’s test. The result: 53% of bootstrapped model subsamples had a larger test distance in ω, with Ω rejectable at 61% and the longitude of perihelion at 62%. Their conclusion was that the sample “is consistent with being detected from a uniform underlying angular distribution”.

Established The Dark Energy Survey reached the same answer twice, on a sample built for cosmology rather than for this question. DES imaged a contiguous 5,000 square degrees of southern sky with about 80,000 exposures in grizY between 2013 and 2019, and its final catalogue holds 817 trans-Neptunian objects, 461 of them new, complete to r ≈ 23.8 over distances of 29 to 2,500 au.

Established Of those, 16 are extreme by the a > 150 au, q > 30 au definition, and the collaboration reports them “consistent with the null hypothesis of azimuthal isotropy”. An earlier four-year analysis with seven extreme objects had already found that “the DES data taken on their own are consistent with azimuthal isotropy and do not require a ‘Planet 9’ hypothesis”. DES also demonstrated something useful in its own right: its detection efficiency between 30 and 2,000 au depends almost entirely on mean apparent magnitude and is independent of light curve amplitude, colour and orbital elements.

Frontier Napier and colleagues then did the thing the argument actually needed: they pooled three characterised surveys and tested the union. Their 2021 paper combined the selection functions of DES, OSSOS and the Sheppard-Trujillo survey over 14 extreme trans-Neptunian objects — five from DES, five from OSSOS, four from Sheppard and Trujillo — and found the individual survey probabilities to be about 0.06 for DES, about 0.53 for OSSOS and about 0.59 for Sheppard-Trujillo, combining to roughly 24%.

Frontier The abstract’s own summary is that the mean scaled longitude of perihelion and orbital poles “are consistent with a uniform population at a level between 17% and 94%”. Remove the single high-inclination outlier 2015 BP519 and the combined figure rises to about 85%.

Frontier That last sentence is the most important number in this brief, and it cuts both ways. A claimed detection whose significance is carried by one object out of fourteen is not a detection; it is a candidate for follow-up. But the same fact undercuts the null: a sample so small that one object moves the p-value from 24% to 85% also has almost no power to exclude a real effect of the size Planet Nine would produce. Neither side is entitled to the conclusion it prefers from 14 objects, and the most defensible reading of the peer-reviewed record is that the characterised data are currently uninformative.

Frontier The two camps are not disagreeing about the data. They are disagreeing about whether an uncharacterised discovery record can be debiased at all. Brown and Batygin build their selection function from the Minor Planet Center’s catalogue of reported discoveries, mapping detectability with a HEALPix equal-area grid and assuming that surveys could have found anything brighter than what they did find. Their stated premise is that “KBOs discovered are a fair representation of the surveys that have been carried out”, and they concede that detailed survey information is unavailable: “With rare exception, such information is not available.”

Frontier They also apply a hand-set correction, weighting the Sheppard et al. 2016 discoveries by a factor of 10 ± 2 to account for that survey’s focus on distant objects. The OSSOS and DES side argues that a selection function you cannot validate is an assumption wearing a statistic’s clothes.

Frontier Brown and Batygin’s counter-argument is equally concrete and equally unresolved. They note that OSSOS’s “limited survey region” renders it “insensitive to the clustering observed in the more evenly distributed surveys”: eight objects from 155 square degrees of deep, narrow pointings sample a small arc of the sky, and a survey that cannot see most of the sky cannot see an all-sky asymmetry. They also chose a conservative inclination width, using σ = 16 degrees, “larger than any of the best fit values” from their own subsets.

Frontier So the honest summary is that one camp discards most of the sample to gain a computable selection function and loses statistical power, while the other keeps the sample and assumes the function. Both trades are defensible and they give opposite answers.

Established Simulation work says plainly that the present sample cannot settle this, and gives a number. Lawler and colleagues pushed Solar System models with and without a distant super-Earth through survey biases and found that “after applying survey biases, the three dynamical models are indistinguishable from each other”, with flux-limited surveys overwhelmingly picking up low-perihelion objects and washing out the signal.

Established Clement and Kaib went further: with 17 objects, the observed clustering sits outside about 98.3% and about 97.7% of random draws in their two measurement spaces, yet “it is not possible to detect the difference between a uniform distribution and a Planet Nine generated distribution at greater than the 1σ level” without roughly 100 or more detections. They also found only 6 of their 17 objects had dynamical lifetimes much longer than the age of the Solar System, which matters because an object that will not survive four billion years is not evidence for a four-billion-year shepherd.

Established Direct searches for the planet have run across the electromagnetic spectrum and found nothing confirmed. A coadded WISE and NEOWISE search of 1,840 square degrees toward Cetus reached W1 < 16.66 at 90% completeness and reported “We do not detect a plausible Planet Nine candidate”. A follow-up 3π search covered 31,480 square degrees at ecliptic latitudes |β| < 55 degrees to W1 < 16.7, again with zero viable candidates.

Established The Atacama Cosmology Telescope ran a blind shift-and-stack search across the whole 18,000 square degrees of its survey, excluded at 95% confidence any unknown Solar System object brighter than 4 to 12 mJy at 150 GHz, listed its ten strongest candidates for follow-up, and eliminated about 17% of the parameter space for a 5-Earth-mass planet and about 9% for a 10-Earth-mass one. “No significant detections are found” is the paper’s own summary.

Speculative Two 2025 far-infrared searches produced candidates, and neither is a planet yet. Phan and colleagues compared IRAS from 1983 with AKARI from 2006 to 2007 — a 23-year baseline — looking for a source that had moved between 42 and 69.6 arcminutes, assuming 7 to 17 Earth masses at 500 to 700 au. Thirteen candidate pairs survived the cuts and one survived image inspection: IRAS F02211-4844 paired with AKARI-MUSL J0220440-491247, separated by 47.4586 arcminutes, with fluxes of 0.24 Jy at 60 µm, 0.27 Jy at 90 µm and 0.52 Jy at 100 µm. Their own verdict: “IRAS and AKARI detections alone are not enough to decide a precise orbit”, and they ask for DECam follow-up.

Speculative A separate AKARI-only search by Chen and colleagues, using sources detected twice within 24 hours and absent after six months, produced two candidates against an expected 0.029 false detections from noise. Candidates are not detections, and no follow-up has confirmed either.

Established Spacecraft ranging constrains where a planet could be without seeing it. Fienga and colleagues used Cassini radio ranging to test a 10-Earth-mass body with a = 700 au and e = 0.6, ruling out true anomalies in the ranges [−130°: −100°] and [−65°:85°] and identifying a most-favourable position at a true anomaly of 117.8 degrees, +11 and −10, where such a planet would actually reduce the Cassini residuals. Across the assumed orbit the distance runs from 280 to 1,120 au. This is a real constraint on a specific orbit family, not a general exclusion, and it depends on the assumed mass and semimajor axis.

Frontier The single most awkward object for the hypothesis arrived in 2025 and was not predicted by anyone. 2017 OF201 has a longitude of perihelion of 306 degrees, which the discovery paper describes as “well outside the clustering region near 60 degrees observed among other extreme TNOs”. Worse for the hypothesis, N-body integrations with two published Planet Nine configurations eject the object in about 60% of realisations on a mean timescale of about 0.3 Gyr.

Frontier The authors state it “lies outside the clustering observed in extreme trans-Neptunian objects, posing a challenge to the proposed dynamical evidence for the hypothetical Planet Nine”.

Established And the instrument that changes the question is already taking data. The Legacy Survey of Space and Time started on 30 June 2026. In roughly six weeks of early commissioning observations Rubin added 380 trans-Neptunian objects to a census of about 5,000 that the community had taken some thirty years to build, including 2025 LS2 and 2025 MX348, whose aphelia reach roughly 1,000 times the Earth-Sun distance. Mario Jurić presented the results at the 248th meeting of the American Astronomical Society on 18 June 2026.

Established The same commissioning data exposed a systematic positional bias of 20 to 40 milliarcseconds in existing asteroid catalogues — a small reminder that the archive everyone has been debiasing against had an uncorrected error in it.

Frontier Question one, and everything else is downstream of it: is the clustering a property of the Solar System or of the observing log? This is not rhetorical framing. The two published debiasing methods give 0.2% and 24% on overlapping data, and the disagreement is methodological rather than numerical. Nothing in the peer-reviewed record resolves it, and both sides have said so in print.

Speculative Question two: if the pattern is real, does it require a planet? A self-gravitating disc of trans-Neptunian objects can do the same work. Sefilian and Touma showed that a disc from 40 to 750 au with the right eccentricity profile produces apsidal confinement because “hot nearly-Keplerian disks induce negative apse precession in their constitutive particles, in contrast to the familiar prograde apse precession expected from cold disks”, cancelling the giant planets’ prograde precession.

Speculative Their required mass runs from about 1 to about 30 Earth masses, with 10 as the fiducial case, and they note the hypothesis “could obviate the need for” a ninth planet entirely.

Speculative The disc route has a live internal problem the field has quantified. Madigan and McCourt’s inclination instability drives exponential growth of orbital inclinations and pericentre clustering from collective gravity alone, and implies an initial disc of about 1 to 10 Earth masses spread between about 50 au and 10,000 au. But Zderic and Madigan then found that the known giant planets suppress it: “differential apsidal precession induced by the giant planets will suppress the inclination instability unless the primordial mass is” at least about 20 Earth masses. Twenty Earth masses of primordial scattered disc is a large thing to require and then lose. The instability does make a falsifiable prediction — a “perihelion gap” at semimajor axes of hundreds of au — which Rubin’s sample can test directly.

Speculative Question three: could the anomaly be in the law rather than in the mass? Brown and Mathur argue that Modified Newtonian Dynamics generates apsidal alignment through the external field effect, with quadrupolar and octupolar terms sourced by the galactic field that have no Newtonian counterpart, on a timescale of 10⁷ years for an object with a semimajor axis of 1,000 au. Their prediction is sharp and testable: “the major axes of the orbits will be aligned with the direction towards the galactic center”.

Frontier They are also unusually candid about the prior, writing that “claimed gravitational anomalies in the solar system have almost invariably proven to be spurious under closer examination”, and they concede that their quadrupole moment “is numerically larger than the bound” from other constraints while remaining of the same order. Dark Matter Alternatives owns the general assessment of modified dynamics; this is one of its few Solar System tests.

Speculative Question four: if there is a mass out there, must it be a planet? Scholtz and Unwin proposed that a captured primordial black hole of roughly 0.5 to 20 Earth masses — motivated by the OGLE microlensing excess — would be dynamically identical to a planet and optically invisible, with a Schwarzschild radius of about 5 cm for a 5-Earth-mass object. Their proposed signature is annihilation in a dark matter microhalo around it, potentially a Fermi-LAT source. Primordial Black Holes covers the mass windows and the abundance limits; what this brief adds is that the Solar System version of the idea has a cheaper test than the gamma-ray one, described in 7.

Frontier Question five is new, quieter, and may be the one that actually resolves first: is there a smaller planet closer in? Siraj, Chyba and Tremaine measured the mean plane of the distant Kuiper belt from 154 non-resonant objects with perihelia beyond 30 au, drawn from a parent set of 740, and found a warp relative to the invariable plane at semimajor axes of 80 to 400 au at 98% confidence and 80 to 200 au at 96% confidence — 2.52σ and 2.74σ, false-alarm probabilities of roughly 4% and 2%.

Speculative Their N-body fit points at an object between Mercury and Earth in mass, around 0.06 to 1 Earth masses, at 100 to 200 au with inclination above about 10 degrees. This “Planet Y” is a different object from Planet Nine, tests a different observable, and is closer, brighter and easier to find.

Speculative Question six: how many perturbers? De la Fuente Marcos and de la Fuente Marcos concluded from Monte Carlo work on the apsidal anti-alignment constraint that “the available data suggest that at least two massive perturbers are present beyond Pluto”. Malhotra, Volk and Wang noted that the four longest-period extreme objects have period ratios near 8/5, 2/1, 8/3 and 10/3 and inferred a planet at about 665 au with a period near 17,117 years. Both results take the clustering as given, so both inherit its uncertainty; they are worth recording because they show how far a small sample can be read, not because they are independent evidence.

Established Bottleneck one is the selection function, and it is not a computing problem or a money problem. To infer the intrinsic distribution of orbital angles you need, for each object, the probability that it would have been found given where every telescope pointed, how deep each pointing went, and which detections were tracked to a secure orbit. OSSOS produced that by injecting millions of synthetic sources into its own discovery images across magnitudes m_r = 21.0 to 25.5. DES produced an equivalent by showing its efficiency depends almost entirely on mean apparent magnitude.

Established For the majority of catalogued extreme objects, no such function exists and none can be reconstructed, because the pointing histories were never published.

Established Bottleneck two: the archive that both sides use as ground truth records discoveries, not searches. The Minor Planet Center holds what was found. It does not hold where people looked and found nothing, which is exactly the quantity a debiasing calculation needs. Brown and Batygin’s method infers the second from the first and says so.

Established Rubin’s commissioning data then found a 20 to 40 milliarcsecond systematic positional bias in existing asteroid catalogues, which is small for orbit determination and a useful demonstration that a heterogeneous archive carries errors nobody had budgeted for.

Frontier Bottleneck three: the sample is small and grows slowly because these objects are faint and far. Fourteen objects carry the published clustering tests. Clement and Kaib put the number needed to separate a uniform distribution from a Planet Nine distribution at more than 1σ at roughly 100 detections. DES, a 5,000-square-degree survey running six years on a 4-metre telescope, produced 16 extreme objects. That is the rate the field has been working at.

Established Bottleneck four: tracking, not discovery. A single-night detection of a distant object is nearly worthless, because the orbit cannot be determined and the object is lost. OSSOS regarded its tracking record — 838 of 840 objects secured, two lost — as its principal achievement, writing that “No previous survey has ever achieved such a high rate of success in tracking TNOs to good orbits.” A survey optimised for transient cosmology does not automatically deliver that, and follow-up time on other telescopes is a scarce resource that nobody has committed at Rubin scale.

Frontier Bottleneck five: the planet’s predicted brightness is not a number, it is a family of assumptions. Brown and Batygin’s own orbit paper declines to give a single magnitude range and instead shows modelled R-band magnitudes varying with albedo and composition, roughly R ≈ 22 to 24 and fainter. The Batygin review a few years earlier put the expected visual magnitude at “of order ~17 or less”, which is a far brighter object.

Frontier The gap between those statements is not a disagreement about physics; it is the absence of any measured albedo for a cold super-Earth at hundreds of astronomical units, and it directly determines how much survey time a search costs.

Speculative Bottleneck six: the most likely hiding place is the worst place to look. The Batygin review notes the “most likely location on the sky where this body could have avoided being discovered to date is the galactic plane” — the region of maximum stellar confusion, where moving-object pipelines are least reliable and where the WISE 3π search cut off at |β| < 55 degrees. A null result over three quarters of the sky is consistent with a planet sitting in the remaining quarter.

Established This subject depends on survey astronomy and on almost nothing else. There is no missing material, no missing detector technology and no missing launch capability in the critical path. Every result quoted in 3 came from a telescope that already existed or a spacecraft flown for another purpose. The dependency is on how surveys are designed, recorded and published.

Established Dependency one: characterised surveys as a methodological standard. OSSOS demonstrated what this costs — artificial source injection at every magnitude, a published efficiency function, near-total tracking — and what it buys: “TNOs discovered with well-quantified discovery biases are key, as they are appropriate to use to understand the intrinsic population distributions.” Every quantitative rebuttal in this brief is downstream of a survey that did that work. Every contested claim is downstream of one that did not.

Frontier Dependency two: survey simulators that outside groups can run. The OSSOS and DES survey simulators are what let Napier and colleagues pool three surveys into one test. Without a distributable simulator, a survey’s selection function stays inside the collaboration and the community gets a catalogue it cannot use for population inference. Rubin’s equivalent is the yield simulation work, which its authors offer so that researchers can “test their methods on an up-to-date, representative, full-scale simulation of LSST data”.

Established Dependency three: spacecraft ranging archives. The Cassini constraint exists because a spacecraft spent thirteen years being tracked to sub-kilometre precision at Saturn for reasons unconnected to this question. Fienga and colleagues describe that dataset as “an exceptional set of measures that acts as a very sensitive device”. Cassini ended in 2017 and nothing replaces it; the next comparable ranging baseline would come from a mission that has not been funded.

Frontier Dependency four: thermal and atmospheric models for objects nobody has observed. Converting a mass to a predicted flux in W1, at 90 µm, or at 150 GHz requires an atmospheric and interior model for a cold ice giant at hundreds of astronomical units. The infrared searches all state their limits in terms of a model — Meisner and colleagues quote sensitivity “at all distances between 250 and ~700 AU” only “for the most optimistic atmospheric model”. The exclusion regions are therefore conditional in a way headline summaries rarely convey.

Established The decisive experiment is already running, and it is the Legacy Survey of Space and Time quadrupling the extreme trans-Neptunian sample under a single characterised selection function. Rubin began the survey on 30 June 2026. Its predicted ten-year solar system yield is about 37,000 trans-Neptunian objects, alongside about 127,000 near-Earth objects, 5.09 million main belt asteroids and about 109,000 Jupiter Trojans, from 1.1 billion observations reaching single-image depths near m_r ≈ 24.0.

Established Roughly 70% of the main belt and the more distant populations are expected in the first two years. The yield paper does not model extreme objects explicitly but states that “they will be discovered early with the other TNOs”.

Frontier What makes it decisive is not the number of objects but that they arrive through one instrument with one cadence. The dispute in 3 exists because 14 objects came from surveys whose selection functions are known to different degrees and in one case were shaped by the hypothesis being tested. A Rubin eTNO sample has a selection function that is computed once and applies to every object in it. That removes the specific degree of freedom the two camps have been exploiting, which is why Konstantin Batygin described the incoming discoveries in June 2026 as a direct test of all lines of evidence for Planet Nine.

Established The sample-size threshold is stated in the literature, so the test has a pass mark rather than a vibe. Clement and Kaib put it at roughly 100 detections to separate uniform from Planet-Nine-generated distributions at better than 1σ. Against a current characterised sample of 14, and a commissioning rate that added 380 trans-Neptunian objects in about six weeks, that threshold is a survey-duration question rather than an instrument question.

Frontier Second experiment: the direct imaging search over the sky Rubin covers. Rubin reaches fainter than the R ≈ 22 to 24 range Brown and Batygin model for the planet under most albedo assumptions, so a detection is possible rather than merely hoped for. The caveat is in their own paper: “the current survey plan does not extend as far north as the full predicted path of Planet Nine”. A Rubin null is a strong constraint over the southern sky and silence elsewhere.

Frontier Third experiment: the Planet Y warp, which is the cheapest falsification on this page. Siraj, Chyba and Tremaine state that Rubin “is expected to confirm or deny the existence of the warp reported here, and might detect the planet that may produce it”. A 2.5σ warp measured on 154 objects either survives contact with tens of thousands or it does not, and no new analysis technique is needed to find out.

Speculative Fourth experiment: follow-up astrometry on the two infrared candidate sets. Phan and colleagues have already named the required observation — DECam imaging of the field around their single IRAS-AKARI pair — and it is a few nights on an existing instrument. If the candidate moves as a bound distant object it is the discovery; if it does not, a much-publicised candidate is retired for the cost of a small proposal.

Speculative Fifth experiment, and the only one that distinguishes a planet from a black hole: accretion flares. Siraj and Loeb argue that if the mass is a black hole, impacts of small Oort cloud bodies produce optical transients that would “be detected at a rate of at least a few per year”, lasting less than the roughly 30-second exposure time, so they appear as point-source flashes within single frames. Rubin’s cadence is the instrument for that search and no separate programme is required.

Handwave Sixth, for completeness: the probe-timing proposal, which is a real calculation attached to hardware nobody has. Witten proposed fleets of laser-launched miniature spacecraft travelling at about 0.001c, reaching the target region roughly a decade after launch, with timing measurements accurate to 10⁻⁵ seconds reported home to detect the gravitational time-delay shift. The arithmetic is sound and the propulsion, the power and the clocks are all unbuilt. It belongs here as an outer bound on what could be done, not as a plan.

Frontier What would change this brief’s assessment fastest is a Rubin null on the clustering itself. A detection settles everything at once and needs no statistics.

Frontier But a Rubin eTNO sample of order 100 objects, debiased with a published selection function, showing longitudes of perihelion consistent with uniform, would retire not just Planet Nine but the disc models and the MOND alignment prediction along with it, since all three exist to explain the same pattern.

Established The engineering that matters here has already been done, and it is a telescope rather than a spacecraft. Rubin’s relevant specification is throughput: a wide field imaged repeatedly to single-image depths near m_r ≈ 24.0, generating 1.1 billion small-body observations over ten years. Mega-Telescopes covers why aperture is priced the way it is; the point for this subject is that the binding parameter is etendue and cadence, not raw aperture, because the target is a moving-object census rather than a single faint image.

Established The hard engineering is in the pipeline, not the optics. Linking detections across nights into orbits at Rubin’s rate is a combinatorial problem on a catalogue of billions of observations, and it is where the selection function is either preserved or destroyed.

Established The commissioning result that exposed a 20 to 40 milliarcsecond positional bias in existing asteroid catalogues came out of that pipeline, which is a fair demonstration that the software is doing real metrology rather than bookkeeping.

Frontier Tracking is the engineering requirement nobody has costed at Rubin scale. OSSOS secured 838 of 840 objects by dedicating observing time to recovery; the ratio of tracking effort to discovery effort was high and deliberate. A survey producing thousands of distant objects a year cannot follow them all the same way, so either the cadence itself has to deliver orbit-quality astrometry or a large share of faint distant discoveries become single-apparition objects that cannot enter a population analysis.

Frontier Infrared search engineering is constrained by archives rather than by new instruments. The strongest infrared limits came from coadding WISE and NEOWISE frames to about 1.2 magnitudes deeper than single exposures, reaching W1 < 16.66 to 16.7. The 2025 candidate work squeezed a 23-year baseline out of IRAS and AKARI, two missions that were not designed for this. There is no funded successor far-infrared all-sky survey, so the archival approach is the available approach.

Speculative Millimetre-wave engineering is a genuine and underused path. The ACT search worked because a cosmic microwave background survey is, incidentally, an all-sky flux monitor with a well-understood beam; it excluded objects brighter than 4 to 12 mJy at 150 GHz across 18,000 square degrees. Next-generation CMB experiments will go deeper over similar areas, and the shift-and-stack technique transfers without modification.

Handwave Anything involving going there is a different engineering problem by two orders of magnitude. Witten’s timing proposal needs sail propulsion at roughly 0.001c, power and clock stability at 10⁻⁵ seconds after a decade in flight, and a communications link from several hundred astronomical units. Planetary Defense documents how long an ordinary deep-space mission takes to authorise and fly for a target that is already catalogued; a target whose existence is disputed does not get to the front of that queue.

Established The nearest adjacent technology is the near-Earth object survey, and it is the same telescope. Rubin’s predicted yield includes about 127,000 near-Earth objects alongside the 37,000 trans-Neptunian ones; the pipeline that links distant slow movers is the pipeline that links fast ones. Planetary Defense owns the impact-risk half of that output. This brief owns the far half, and both depend on the same cadence decisions.

Frontier Cosmic microwave background surveys are adjacent in an unobvious way. A CMB experiment is an all-sky, repeated, absolutely calibrated flux measurement, which makes it a moving-object survey if you are willing to shift and stack. ACT covered 18,000 square degrees and produced Solar System limits as a by-product of measuring the early universe.

Frontier Precision spacecraft ranging and planetary ephemerides are adjacent and constrain the same parameter space. The Cassini result rules out ranges of true anomaly for a specified 10-Earth-mass orbit. Any future outer Solar System mission tracked to comparable precision inherits this capability free, which is a real argument for range-capable transponders on missions flown for other reasons.

Speculative Dark matter astrophysics is adjacent through two separate doors. If the perturber is a primordial black hole, the search becomes a dark matter search with a Fermi-LAT signature, which is Primordial Black Holes’ subject. If the anomaly is gravitational rather than material, it becomes a test of modified dynamics in the deep weak-field regime, which is Dark Matter Alternatives’ subject. Neither door is opened by anything this brief can do on its own.

Established Archival data mining is adjacent and has produced results this year. 2017 OF201 was found by a systematic search for moving objects in Dark Energy Camera Legacy Survey archives, with ten detections in g, r and z between 2014 and 2018 and additional observations eventually spanning 2004 to 2025. The same technique produced the IRAS-AKARI candidate. Archives are cheap and the objects were already in them.

Established The central institutional fact is that the Minor Planet Center records discoveries and not searches, and every debiasing dispute in this subject traces back to that. A registry of what was found cannot answer what would have been found. Brown and Batygin’s method infers survey coverage from the discovery record and says so explicitly: “With rare exception, such information is not available.” No institution is tasked with collecting the missing half.

Frontier A survey’s selection function is currently a courtesy, not a requirement. OSSOS and DES published theirs and the field’s strongest results follow from that choice. There is no journal policy, funding condition or archive standard obliging a survey to publish an efficiency function or a runnable simulator, which means the analysability of a dataset depends on the habits of the collaboration that produced it.

Frontier The field has a documented incentive problem in its pointing strategy. Brown and Batygin apply a weighting factor of 10 ± 2 to the Sheppard et al. 2016 discoveries specifically because that survey targeted distant objects, and Sheppard and Trujillo were explicit that their programme was searching “towards a super-Earth in the outer solar system”.

Frontier Searching where a hypothesis predicts objects is good observing practice and terrible statistics, and the correction for it is a hand-set number with an uncertainty attached by judgement.

Speculative Nothing in this field is pre-registered. The sample cut that defines an “extreme” object has moved between a > 150 au, a > 230 au and a > 250 au across the papers cited here, and the choice of statistical test has varied between Kuiper’s test and bespoke Monte Carlo procedures. Those are all defensible choices made after seeing the data. A pre-registered analysis plan for the Rubin sample, agreed before the objects arrive, would cost nothing and would remove the one criticism that both camps can currently level at each other.

Established Funding for this question is entirely derivative and that is mostly a strength. Rubin was built for cosmology and time-domain astronomy; DES was built to measure dark energy; ACT was built to measure the microwave background; Cassini went to Saturn. Every constraint in 3 is a by-product. The corresponding weakness is that no instrument has ever been optimised for this target, so the null results are nulls over the parameter space someone else’s cadence happened to cover.

Frontier The live ethical issue in this subject is communication, not risk. No one is harmed by an undiscovered planet. But “Planet Nine” has had a decade of coverage in which a contested statistical inference is routinely reported as a near-certainty, and the same field published, in peer-reviewed journals, that its characterised samples are “consistent with a uniform population at a level between 17% and 94%”. Both statements came from the same discipline and the public heard one of them.

Established The naming of candidates carries real reputational cost. Two far-infrared searches in 2025 announced candidates that their own authors described as needing follow-up to establish an orbit, and both were reported widely as possible detections. Phan and colleagues wrote that their detections “are not enough to decide a precise orbit”. A field that produces candidates faster than it retires them accumulates a backlog that looks, from outside, like evidence.

Frontier There is an allocation question hiding behind the popular interest. Telescope time is rationed by committee, and a hypothesis with a large public profile competes for follow-up against proposals with better-defined priors. The honest position is that the follow-up costs discussed in 7 are small — a few nights of DECam — and that this is precisely why the backlog of unretired candidates is hard to excuse.

Speculative If a planet is found, the naming fight will be disproportionate to the science. The International Astronomical Union’s 2006 redefinition of “planet” remains contested in public, and a 6-Earth-mass body at 380 au would be an unambiguous planet by every criterion, discovered by a survey rather than an individual. That is a small problem and it is worth naming so that it does not get confused with a large one.

Speculative A confirmed ninth planet changes the inventory of the Solar System more than it changes anything humans do. It would be the first planet added since Neptune in 1846, it would be the nearest example of the super-Earth or mini-Neptune class that dominates exoplanet catalogues, and it would sit roughly thirteen times further out than Neptune. None of that alters a budget, a treaty or an industry.

Frontier The durable civilizational product is the census, not the planet. About 37,000 trans-Neptunian objects with computed detection probabilities is a structural map of the outer Solar System, and structural maps of that kind constrain formation models, migration histories and the origin of the objects that occasionally cross Earth’s orbit. That output arrives whether or not a planet is in it.

Speculative A detection would put a real target at a distance that is currently rhetorical. Missions to 300 to 800 au are discussed in the abstract; a confirmed body with an ephemeris converts an abstract destination into a specific one, which historically is the difference between a concept study and a programme. Nothing in the record says such a mission would be funded, and 8 says why the propulsion case is unbuilt.

Frontier The most transferable lesson is methodological and applies well outside astronomy. This is a clean, public, decade-long case of a discipline discovering that its headline result depended on a selection function nobody had measured, and then building the instrument that measures it. Fields with observational rather than experimental data — epidemiology, economics, palaeontology — have the same structural problem and rarely get to watch it resolved.

These horizons track three separate clocks — the statistical question, the direct search, and the alternatives — because only the first has a schedule:

• 10 yr: Established Rubin’s ten-year survey, started 30 June 2026, delivers about 37,000 trans-Neptunian objects with roughly 70% of the distant populations in the first two years. Frontier The clustering question should be answered within that window, because the characterised sample passes the roughly 100 detections Clement and Kaib say are needed well before the survey ends. Frontier The Planet Y warp at 80 to 200 au is confirmed or denied on a shorter clock still, since the objects involved are brighter. Speculative The two 2025 infrared candidates are either recovered by DECam follow-up or quietly dropped; nothing forces a resolution. Frontier Direct detection, if the planet exists and lies in Rubin’s sky, is most likely in this decade — but the survey “does not extend as far north as the full predicted path”, so a null here is partial.
• 25 yr: Frontier If the clustering survives debiasing and no planet is imaged, this is the window in which the disc and modified-gravity alternatives get tested on their own predictions — the perihelion gap for the inclination instability, alignment with the galactic centre for the external field effect. Speculative A next-generation microwave background survey extends the 4 to 12 mJy limits at 150 GHz over comparable sky. Speculative A far-infrared all-sky successor to AKARI would close the archival gap; none is funded.
• 50 yr: Speculative If a body is confirmed, this is the earliest plausible window for a flyby mission, and it inherits the propulsion and authorisation problems that Planetary Defense documents for far nearer targets. Speculative If nothing is confirmed, the subject becomes a chapter in the selection-bias literature rather than a planetary science programme.
• 100 / 250+ yr: Handwave Witten’s timing-fleet architecture — laser-launched craft at about 0.001c with 10⁻⁵-second clocks — is the only proposal that finds a mass regardless of whether it emits light, and every component of it is unbuilt. Handwave On these timescales the distinction between a 6-Earth-mass planet and a 5-centimetre black hole becomes an engineering specification rather than a curiosity, which is the one sense in which the exotic alternative matters practically.

• Depends on Nothing on this map blocks this subject, and the absence is worth stating plainly rather than padding. No material, detector, launcher or theoretical result in another brief stands between the field and an answer. The shared instrument is Rubin, whose near-Earth output Planetary Defense covers and whose cost-versus-aperture logic Mega-Telescopes owns, but neither is a dependency: this subject would proceed at a slower rate on 4-metre survey telescopes exactly as it did for the previous decade.
• Requires (not on this map) Five things, and none is a discovery. One, a selection function covering the whole known extreme sample rather than the fraction from characterised surveys: today the 14 objects in the pooled test come from three programmes and the remainder cannot be analysed at all. Two, an archive that records where telescopes looked and saw nothing; the Minor Planet Center records only discoveries, which is why both camps must infer coverage instead of reading it. Three, committed follow-up time to track faint distant discoveries to secure orbits — OSSOS regarded its 838-of-840 tracking rate as its main achievement, and nothing guarantees that rate at Rubin volumes. Four, a measured albedo and atmospheric model for a cold super-Earth beyond 300 au, without which the predicted magnitude spans from about 17 to fainter than 24 and every exclusion limit is model-conditional. Five, an analysis protocol agreed before the Rubin sample arrives, given that the extreme-object cut has already moved between 150, 230 and 250 au across the published papers.
• Enables A calibrated outer Solar System census enables population-level tests of planet formation and migration that no current dataset supports, and it is a prerequisite for any serious proposal to send hardware beyond a few hundred astronomical units. A confirmed compact perturber would also hand Primordial Black Holes its first Solar System test case.
• Adjacent Planetary Defense shares the instrument, the pipeline and the tracking problem, and differs entirely in what it does with the output. Mega-Telescopes owns the cost structure of the aperture this subject borrows. Dark Matter Alternatives owns modified dynamics, of which the external field effect explanation for apsidal alignment is a Solar System test. Primordial Black Holes owns the mass windows that make a 5-centimetre perturber a live idea rather than a joke. Outside this map: survey astronomy pipelines, planetary ephemerides, and the statistics of small-sample circular data.

Frontier “Planet Nine has been detected at 3.8 sigma.” No. The 3.8-sigma figure is the 2016 joint probability of 0.007% for six objects, and it has not survived the addition of data. The same authors reported 1.2% on ten objects in 2017, 0.2% on fourteen objects in 2019, and 99.6% confidence in 2021. Those are not independent confirmations; they are successive restatements of one claim, and the trend is toward weaker significance as the sample grows.

Frontier A sigma value attached to an inference about a selection function is not the same kind of object as a sigma value attached to a particle excess.

Established “The rebuttals are just scepticism.” They are quantified results using published survey simulators, and they were peer reviewed. Shankman and colleagues ran Kuiper’s test on the OSSOS selection function and reported rejectability at 53%, 61% and 62% for the three angles. Bernardinelli and colleagues reported the DES extreme objects “consistent with the null hypothesis of azimuthal isotropy” in two separate data releases. Napier and colleagues pooled three surveys and got 17% to 94%. Anyone characterising this literature as hand-waving has not read it.

Frontier “So the clustering has been disproved.” Also no, and this is the symmetric error. The characterised samples are tiny. Napier’s combined probability moves from about 24% to about 85% on the removal of a single object, 2015 BP519, which means the test has very little power in either direction. Lawler and colleagues showed that after survey biases are applied, models with and without a distant planet are “indistinguishable from each other”. A non-detection of clustering in a sample that could not have detected it is not evidence of absence.

Established “Surveys have already ruled out a planet.” They have ruled out parts of parameter space, and the parts are smaller than the headlines. The ACT search, covering 18,000 square degrees, eliminated about 17% of the parameter space for a 5-Earth-mass planet and about 9% for a 10-Earth-mass one. The WISE 3π search covered 31,480 square degrees but only at |β| < 55 degrees, and the review literature says the most likely hiding place is the galactic plane.

Established The Cassini constraint excludes two ranges of true anomaly for one assumed orbit family, not the planet.

Speculative “The 2025 infrared candidate is probably it.” The discovery paper says otherwise in its own words: IRAS and AKARI detections “are not enough to decide a precise orbit”. A single pair of positions 23 years apart constrains a direction of motion, not an orbit, and the required DECam follow-up has not been reported. A second, separate AKARI-only search produced two further candidates against an expected 0.029 noise false detections. Multiple unconfirmed candidates from different methods is what a false-positive population looks like as well as what a real object looks like.

Speculative “If it is not a planet it must be a black hole.” The primordial black hole version is a legitimate paper — Scholtz and Unwin, 0.5 to 20 Earth masses, a 5-centimetre Schwarzschild radius for 5 Earth masses — and it is strictly an alternative for the composition of a perturber whose existence is the actual question. It also inherits a separate burden: it needs primordial black holes to exist in that mass window in the required abundance and to be captured, which Primordial Black Holes covers and does not concede.

Frontier “Planet Nine and Planet Y are the same claim.” They are not, and conflating them inflates both. Planet Nine is inferred from apsidal and pole clustering in objects beyond 230 to 250 au and is modelled at 5 to 10 Earth masses with a semimajor axis of hundreds of astronomical units. Planet Y is inferred from a warp of the mean plane at 80 to 200 au, at 2.52σ and 2.74σ on 154 objects, and is modelled at 0.06 to 1 Earth masses. Different data, different observable, different object. Either could be true without the other.

Established “A ninth planet would overturn planetary science.” It would be a major discovery and an ordinary one. Super-Earths are the commonest known exoplanet class; a captured or scattered ice giant is a standard outcome in formation models; and the Solar System’s known architecture already accommodates objects like Sedna that nothing currently explains. The genuinely unusual outcome would be the other one — establishing that a decade of headline statistics was a map of telescope pointings — which is the outcome this brief treats as roughly as likely.

Handwave “There are at least two perturbers out there.” That is a real published conclusion — de la Fuente Marcos and de la Fuente Marcos, from Monte Carlo work on the anti-alignment constraint — and it is conditional on the clustering being real, on the specific objects used, and on the dynamical model. Multiplying hypothetical planets to fit a fourteen-object sample is where this subject is least disciplined, and it is the reason the flag on that claim is the weakest one available.