Dark Matter Alternatives

Speculative If the acceleration scale that appears in galaxy rotation curves is a property of the gravitational law rather than an accident of how dark haloes assemble — which no experiment has shown, and section 7 says what would — then roughly a quarter of the mass-energy budget of the universe is a bookkeeping artefact, and a number near 1.2 × 10⁻¹⁰ m s⁻² joins G and c as a constant to be measured rather than a parameter to be fitted. The conditional is the whole of it: that sentence is a hypothesis with one striking correlation behind it and two unmet obligations in front of it.

Established The magnitude is best stated as the price of the null, because the null is the only thing either programme has actually delivered. The LUX-ZEPLIN detector ran 417 live days between March 2023 and April 2025 on about ten tonnes of liquid xenon, reported no dark-matter signal, and is about three million times more sensitive than the instruments its spokesperson used when he entered the field. XENONnT excluded spin-independent dark-matter–nucleon cross-sections above 2.5 × 10⁻⁴⁵ cm² at 6 GeV/c² and 90% confidence. The cost of the next step is now measurable: a 93% increase in exposure bought a 10% improvement in median sensitivity at 5 GeV/c², because solar neutrinos have started to look like the signal. That exchange rate, not any theoretical argument, is what gives the alternatives a hearing.

Frontier Who collects the upside is unusually lopsided, and the asymmetry is in the cost of being wrong rather than in the physics. The particle programme is capital: tonne-scale radiopure xenon, deep underground halls, collaborations of several hundred people, and a decade between decisions. The modified-dynamics programme is archival: the results in section 3 come from public catalogues — rotation curves from a Spitzer photometric database, wide binaries from Gaia — analysed by groups of two or three at effectively no marginal cost. One side can be wrong for forty years and still be employed; the other can be wrong in a single paper. Nothing at the end of either is a product, and this page does not pretend otherwise.

Handwave The condition, stated plainly: the upside is real only if one of the two programmes is falsifiable on a schedule, and at the moment neither is. The particle side responds to every null by lowering the assumed cross-section, which is legitimate and unbounded; the modified-dynamics side responds to every failure by adding a component — a sterile neutrino, a residual collisionless mass in clusters — that is dark matter under another name. Section 5 is about whether either retreat has a floor.

Established The observation everything rests on is that gravitating systems larger than a star cluster move faster than their visible contents can explain. Spiral galaxies rotate with flat outer velocity curves; clusters of galaxies hold together at velocities their hot gas and stars cannot account for; the cosmic microwave background and the growth of structure both want about five times more gravitating matter than the baryon density inferred from primordial nucleosynthesis. Every serious position in this argument accepts these measurements. The dispute is about what to add.

Established The standard answer adds a substance. Cold dark matter is a collisionless, non-baryonic particle species that clusters gravitationally and interacts weakly or not at all otherwise. It is the assumption that makes the concordance cosmology work, and the search for its particle is treated as an atlas of excluded parameter space in Hidden-Sector Searches. That brief maps where the particle is not. This one asks the prior question: whether there is a particle.

Frontier The principal alternative changes the law instead. Modified Newtonian dynamics, proposed by Milgrom in 1983, posits that below an acceleration scale a₀ the relation between gravitational field and source departs from the Newtonian one, so that in the deep-low-acceleration limit the effective acceleration goes as the geometric mean of the Newtonian value and a₀. Because orbital speed then depends only on enclosed baryonic mass, flat rotation curves and a tight mass–velocity relation follow without a halo. The theory has one new constant and one free interpolating function, and the second of those is doing more work than its advocates usually concede.

Speculative A third position denies the dichotomy: the missing mass could be neither a new particle nor a new law, but ordinary matter in a form telescopes do not see. Primordial black holes are the surviving version of this idea, and they are dark matter that is not a particle — which makes them a genuine third option rather than a variant of the first. Handwave Every non-particle baryonic candidate large enough to matter has been squeezed by microlensing and by the same nucleosynthesis budget that created the problem, and the remaining windows are narrow enough that this brief treats the option as live but unpromising.

Speculative What separates these three is not only evidence but what each side will accept as an explanation. A fit with a free function is, to one camp, a law discovered and, to the other, a curve drawn. A simulation with tuned feedback prescriptions is, to one camp, physics and, to the other, an epicycle. The argument in section 4 is partly empirical and partly about the standard of explanation, and pretending otherwise is how this subject produces forty years of papers that do not engage each other. Frontier Philosophy of Science and Scientific Revolutions treat that structure generally; this brief treats the one case where the disagreement has numbers attached.

Established The measured anchor of the modified-dynamics case is the radial acceleration relation, and it is a real and impressive correlation. McGaugh, Lelli and Schombert reported in Physical Review Letters in 2016 that the acceleration observed at each radius in a rotation curve is a tight function of the acceleration predicted by the baryons alone, with 2,693 data points in 153 galaxies of very different morphology, mass and gas fraction falling on one curve. The fitted scale is g^† = 1.20 ± 0.02 (random) ± 0.24 (systematic) × 10⁻¹⁰ m s⁻², the observed root-mean-square scatter is 0.13 dex against an expected observational scatter of 0.12 dex, and the authors describe the data as consistent with negligible intrinsic scatter.

Established Three corrections to how that result is usually quoted, because this brief was commissioned against a version of it that compressed all three away. First, the paper reports 2,693 points in 153 galaxies, not 153 rotation curves each reduced to a point; the sample is late-type disc galaxies, and the extension to 240 objects adding 25 early-type galaxies and 62 dwarf spheroidals is a separate 2017 paper. Second, the systematic uncertainty on the scale is ±20%, an order of magnitude larger than the random error and almost always dropped in quotation. Third, and most important, the paper writes g^† rather than a₀ on purpose: identifying the empirical scale with Milgrom’s constant is a theoretical step, not a measurement.

Frontier Whether the scale is universal at all is contested, and the contest is a methodological fault line rather than a data problem. Li, Lelli, McGaugh and Schombert fitted the relation to 175 individual SPARC galaxies by Markov chain Monte Carlo, marginalising over stellar mass-to-light ratio, distance and inclination, found residual scatter of 0.057 dex (about 13%), and reported that allowing the critical acceleration to vary from galaxy to galaxy did not improve the fits: “there is no credible indication of variation in the critical acceleration scale”. Frontier Rodrigues, Marra, Del Popolo and Davari, analysing 193 disc galaxies from the same public databases by Bayesian inference, rejected a universal acceleration scale at more than 10 standard deviations with a p-value below 10⁻²⁰ and concluded the scale is emergent. Established Two Bayesian analyses of overlapping public data reached opposite conclusions about the central claim of the field, and the published exchange that followed turned on prior choices for nuisance parameters, chi-squared quality cuts and how posterior compatibility is assessed — not on any new observation. That pattern recurs in section 7 and is the reason its decisive experiment is a protocol rather than an instrument.

Established The strongest single result against pure modified dynamics is the Bullet Cluster, and it is strong for a reason often stated badly. Clowe and colleagues reported weak-lensing observations of the merging cluster 1E0657-558 at redshift 0.296 in which the reconstructed gravitational potential peaks are spatially offset from the X-ray plasma — by roughly 7 arcseconds for the subcluster — at about 8 standard deviations. The plasma, not the galaxies, is the dominant baryonic component, so any force law that scales with the baryons should centre the lensing signal on the gas. It does not. The argument is about the location of the mass, not its amount.

Frontier MOND’s own literature does not agree with itself about the answer, and both 2026 papers are worth reading. Famaey, analysing updated lensing models informed by JWST, found that within 300 kpc of the main and secondary clumps MOND predicts projected mass ratios of about 2.8 and 3.4 times the baryonic mass while the observations require about 7.5 and 9 — a residual missing mass of 3.4 × 10¹⁴ solar masses, comparable to all the baryons present, which must be mostly collisionless and concentrated near the galaxies. Frontier Hernandez, in the same year, argued that the offset itself is not the problem: in QUMOND the anomaly depends on volume density rather than total mass, so the compact galaxies produce 48% of the phantom mass while holding only about 7% of the baryons, and the predicted lensing signal peaks on the galaxies. Established Taken together these say that the offset is survivable and the deficit is not: MOND in clusters still misses a factor of two to three, which its own principal advocate’s public test table has recorded for years.

Established The cosmic microwave background produced the cleanest falsification in this subject, and it was conceded by the person who made the prediction. A baryons-only model with no cold dark matter predicted the amplitude ratio of the first to second acoustic peaks correctly and in advance, in 1999, when the concordance expectation did not. The same model predicted a third peak far lower than the one subsequently measured. McGaugh’s own statement is that no-CDM “fails to correctly predict the second-to-third peak amplitude… It is, in fact, way off, while LCDM is consistent” with it. Speculative A relativistic theory published by Skordis and Zlosnik in Physical Review Letters in 2021 is cited by the same author as the reply — a modified-gravity construction that reproduces the observed spectrum — and this brief could not retrieve that paper, so it states the attribution and asserts nothing further about what the theory achieves.

Established The one MOND-based cosmology with published structure-formation simulations fits the microwave background only by paying a price elsewhere, and the price has now been measured. The νHDM model keeps Milgromian gravity and replaces cold dark matter with a sterile neutrino of about 11 eV. Fitting it to the angular power spectrum, Samaras, Grandis and Kroupa obtained H₀ = 55.64 ± 0.32 km s⁻¹ Mpc⁻¹ and Ω_m = 0.495 ± 0.009, against 67.36 and 0.315 for the standard model — a Hubble constant far below every local and early-universe determination rather than between them. Frontier That direction matters, because modified dynamics is sometimes advertised as relieving the expansion-rate discrepancy: the local and early-universe determinations sit between about 67 and 73, and Cosmological Tensions treats that gap in detail. A model landing at 56 has not split the difference; it has left the interval.

Established The largest MOND cosmological simulation yet run excludes that model at more than five standard deviations. Russell, Banik, Cray and Zhao simulated an 800 h⁻¹ Mpc box with 256³ particles from redshift 199 and report that νHDM massively overproduces large-scale structure: the most massive cluster reaches about 5 × 10¹⁷ h⁻¹ solar masses, typical peculiar velocities are several thousand kilometres per second against hundreds in the standard model, and the resulting bulk flows exceed the Cosmicflows-4 measurements, ruling the model out at more than 5σ. Their conclusion is broader than MOND: replacing cold dark matter with hot dark matter is unlikely to give a viable cosmology regardless of the gravity law.

Established On the particle side the record is a deepening null that has now reached an irreducible background. LUX-ZEPLIN reported 417 live days over two years with world-leading sensitivity above 5 GeV/c² and the first xenon search below 9, and detected solar boron-8 neutrinos at 4.5σ. XENONnT measured the same coherent neutrino–nucleus scattering at 3.3σ in 6.77 tonne-years, inferring a boron-8 flux of (5⁺³₋₂) × 10⁶ cm⁻² s⁻¹, and stated explicitly that sensitivity now shows diminishing returns with exposure. Established The quantitative form of that statement is the most consequential number in this brief: a 93% increase in exposure improved median sensitivity at 5 GeV/c² by 10%.

Frontier The small-scale problems that motivate the alternatives are real, partly addressed, and contested on both sides. Bullock and Boylan-Kolchin’s review sets out the arithmetic: simulations predict as many as a thousand subhaloes above 10⁷ solar masses around a Milky Way against roughly fifty observed satellites; dwarf central densities come out about 50% below dissipationless expectations; and supernova feedback produces cores only above a stellar-to-halo mass ratio of about 10⁻⁴, which leaves the classical and ultra-faint dwarfs unexplained. Frontier Feedback has since absorbed much of the missing-satellite and core-cusp problem; the diversity of rotation-curve shapes at fixed circular velocity, coexisting with the tightness of the radial acceleration relation, has not been absorbed and is the version of the problem that survives.

Frontier Attempts to derive the relation inside the standard model exist and are partial. Tian and Ko showed that re-expressing the data as a halo acceleration relation — the dark component’s own acceleration against the baryonic one — discriminates between functional forms that fit the radial relation equally well, and that both cuspy and cored halo profiles have difficulty with it. Speculative Self-interacting dark matter is the standard-model response that addresses diversity directly, by letting halo cores respond to the baryons; the 2019 Physical Review X paper usually cited for it could not be retrieved for this brief, so its parameters are not quoted here.

Frontier Warm dark matter, the other intermediate position, is not currently constrained by galaxy scaling relations. The 2026 CASCO analysis of the DREAMS hydrodynamic simulations recovered a warm-particle mass of 2.7^+1.3_−0.6 keV from SPARC observations while noting that in uniform-box runs the particle mass does not significantly alter the trends at all; the one signature they judge separable from feedback is a suppression of galaxy abundance below a stellar mass of 10⁸ solar masses.

Frontier A dynamical test at the smallest scales currently favours Newton. Dynamical friction acts differently without a halo, and Bilek, Combes, Nagesh and Hilker simulated globular clusters sinking in gas-rich dwarfs under both laws. Clusters below about 10⁵ solar masses survive in either, held up by supernova-driven potential fluctuations; above about 4 × 10⁵ solar masses feedback cannot stop them sinking in MOND, producing offset-core galaxies that are not observed. The authors describe the result as favouring Newtonian gravity, and they also describe it as dependent on the accuracy of the feedback model.

Frontier Is the acceleration scale fundamental or emergent? This is the load-bearing question and it is currently answered both ways in refereed journals from the same data. If the scale is fundamental, it is a constant of nature and the dark-halo interpretation is an elaborate coincidence; if it is emergent, it is a consequence of how baryons settle into haloes and the correlation is a result to be derived rather than a law to be adopted. Speculative No observation currently distinguishes the two, because a sufficiently flexible galaxy-formation model can produce a tight relation and a sufficiently flexible interpolating function can produce any curve.

Frontier Can a relativistic completion of MOND reproduce the third acoustic peak without a component that behaves like dark matter? This is the sharpest form of the cosmological objection, and the honest statement of its status is uncomfortable for both sides. The baryons-only prediction failed. The relativistic constructions that succeed do so by introducing fields whose energy density behaves, at the relevant epoch, like pressureless matter. Handwave Whether a field that mimics dark matter in the early universe and modifies dynamics in galaxies is a unification or a relabelling is a question this brief cannot settle and neither camp asks in public.

Frontier Is there a gravitational anomaly in Gaia wide binaries below a₀? Wide binary stars at separations beyond about 0.01 parsec probe accelerations below the MOND scale in a system with no plausible dark halo of its own, which makes them the cleanest available test. Hernandez and colleagues report a velocity-boost parameter of 1.5 ± 0.2 in the low-acceleration regime against 1.000 ± 0.096 at high acceleration; Chae and Yoon report a boost of 1.3 to 1.6 at more than 5σ below 10⁻⁹ m s⁻². Frontier Other groups analysing the same catalogue report consistency with Newton, and the disagreement is entirely about sample selection and the modelling of undetected triple systems. Speculative A field in which two groups extract more than 5σ in opposite directions from one public dataset has a methodology problem rather than a data problem, and the methodology problem is fixable.

Frontier Does MOND’s cluster residual have a baryonic form? The factor of two to three missing in clusters has been attributed to cold gas clouds, to faint compact objects, and to sterile neutrinos. Speculative Famaey’s 2026 analysis requires the residual to be mostly collisionless and to track the galaxies rather than the gas, which rules out the simplest gas-cloud version and pushes the candidate toward something that behaves exactly as dark matter does. Handwave At that point the modified-dynamics position has conceded the substance and is arguing about its abundance, which is a smaller claim than the one the field advertises.

Speculative Could the missing mass be non-particulate? Primordial black holes remain the only candidate that is neither a new particle nor a new law. Their appeal is that they require no new physics beyond an early-universe perturbation spectrum, and their difficulty is that microlensing surveys, the microwave background and the gravitational-wave merger population each close different mass windows. Speculative The asteroid-mass window that survives is the one hardest to probe, which is an uncomfortable position for a hypothesis to occupy and not by itself an argument against it.

Frontier Which anomalies are the theory’s and which are the simulation’s? A large share of this dispute is about whether a discrepancy between observation and a simulation is evidence against the physics or against the subgrid prescription. Established The globular-cluster result is explicit about this: it favours Newtonian gravity and says in the same breath that the conclusion depends on the feedback model. Speculative No agreed procedure exists for deciding when a feedback prescription has been tuned past the point of testability, and inventing one would do more for this subject than another survey.

Speculative Does the external field effect exist? MOND predicts that the internal dynamics of a system depend on the external gravitational field it sits in — a violation of the strong equivalence principle with no counterpart in any dark-matter model. Speculative It is the one prediction that is uniquely MOND’s rather than shared, and claimed detections in dwarf galaxy velocity dispersions and in rotation-curve declines are at the edge of the systematics. Handwave A clean external-field-effect measurement would be worth more to this field than any further rotation curve, and nobody has made one that the other side accepts.

Frontier The first bottleneck is the interpolating function, which is a free function standing where a derivation should be. MOND specifies behaviour in two asymptotic limits and says nothing about the transition between them, and the transition region is where most rotation-curve data live. Different choices change fitted parameters by amounts comparable to the effects under test, and the choice is made by fit quality rather than by principle. Handwave Until the function is derived from something, a MOND fit is a two-parameter description with one parameter hidden in the shape.

Established The second is stellar mass-to-light ratio, the systematic that dominates every rotation-curve result. Converting light to mass requires a stellar population model, and the ±20% systematic on the radial acceleration relation’s scale is largely this. Frontier It is also why two groups can fit the same curves and disagree: marginalising over mass-to-light with different priors is precisely the step on which the universal-scale dispute turned. Speculative Asteroseismology and resolved-population calibration could cut this, and no programme is funded to do it at the scale a thousand galaxies would need.

Established The third is the neutrino background, and unlike the others it is physics rather than technique. Coherent scattering of solar boron-8 neutrinos produces nuclear recoils indistinguishable event-by-event from a light dark-matter signal, and both leading xenon experiments have now measured it — 4.5σ and 3.3σ. The consequence is the exchange rate already quoted: 93% more exposure for 10% more reach at 5 GeV/c². Speculative Whether this is a floor or a slope is a real and separate question: directional sensitivity and annual modulation carry information that raw exposure does not, and this brief read no source that settles it, so it does not claim the fog is impassable.

Frontier The fourth is computational asymmetry. Cold-dark-matter cosmology has three decades of simulation infrastructure, calibrated feedback libraries and a workforce; MOND cosmology has a handful of codes and the largest published box is 800 h⁻¹ Mpc with 256³ particles. Speculative That asymmetry means a MOND prediction is usually cruder than the standard-model prediction it is compared against, and a crude prediction that fails is weaker evidence than a careful one that fails. Handwave It also means the asymmetry is self-reinforcing, because the crude failure is then cited as a reason not to fund the careful version.

Frontier The fifth is that the wide-binary test is limited by things Gaia was not built to measure. Undetected tertiary companions inflate relative velocities in exactly the way a gravity boost would; the eccentricity distribution of wide pairs is unmeasured and is an input to every analysis; and the quality cuts that remove contaminants also remove the most precisely measured systems. Speculative Chae and Yoon argue that one widely used error cut on normalised velocity “biases the data in a way to selectively remove accurate and valid velocities”, which is a claim about a competitor’s procedure rather than about the sky, and it is the kind of claim a pre-registered protocol exists to retire.

Frontier The wide-binary test depends on Gaia DR4 and on nothing else in this list. The present dispute is fought over DR3 astrometry; a release with longer time baselines and published astrometric binary solutions would identify a large share of the contaminating triples directly rather than modelling them statistically, which is the single assumption the disagreement turns on.

Frontier The cluster question depends on lensing reconstructions good enough to separate a residual that tracks galaxies from one that tracks gas. Famaey’s 2026 result used JWST-informed models, and the substantive claim — that the residual is collisionless and galaxy-centred — is a statement about where mass sits at the 100 kpc scale. Speculative A sample of merging clusters rather than one famous example is what would make it a measurement rather than a case study.

Established The stellar mass-to-light calibration is an input from a different discipline. Every number in section 3 that involves baryonic mass inherits the uncertainties of stellar population synthesis, and the largest single systematic in the field is therefore owned by people who are not in it.

Frontier The cosmological arm depends on the growth-of-structure and lensing data that the current generation of surveys is producing. A modified-gravity theory that fits the microwave background must also fit the late-time growth rate and the lensing amplitude, and these are the measurements that discriminate between constructions with different auxiliary fields. Frontier Multi-Messenger Astronomy covers the instrument side of this; the constraint that gravitational waves and light arrive together, measured to a part in 10¹⁵, already eliminated a family of relativistic modified-gravity theories, which is the clearest example in this subject of a decisive result arriving from a neighbouring field.

Speculative The particle arm depends on supply chains rather than ideas. The limiting inputs for the next generation are radiopure xenon and argon at ten to fifty tonnes, underground space, and the low-background fabrication described in Quantum Materials; none of these is a physics problem and all of them are decade-scale commitments.

Frontier Every position above has a test, and they rank by how much a result would move the argument rather than by cost.

Speculative Tier one, and the decisive experiment for this brief: the Gaia wide-binary test below a₀, run under a selection and triple-contamination protocol that both camps register jointly before anyone unblinds. The reason this ranks first is not that wide binaries are the most important system but that they are the only one where a definite answer is currently being prevented by something a committee could fix. Established The data already exist and have already produced more than 5σ in both directions — a boost parameter of 1.3 to 1.6 below 10⁻⁹ m s⁻² from one group, consistency with Newton from another, on overlapping Gaia samples. Frontier The protocol would have to fix, in advance: the acceleration bins, the renormalised unit weight error and radial-velocity cuts, the triple-system model and the sub-sample used to calibrate its fraction, the assumed eccentricity distribution, and the statistic reported. Handwave No such protocol has been agreed, nobody is obliged to agree one, and the cheapest decisive experiment in this brief is therefore also the one least likely to be run.

Speculative Tier two is the cluster residual, measured on a sample rather than on the Bullet Cluster. The question is sharp: in merging clusters, does the mass that neither gas nor stars account for follow the galaxies or the plasma? Frontier Famaey’s analysis says galaxies, and requires 3.4 × 10¹⁴ solar masses of it in one system. Speculative If a sample of ten to twenty merging clusters shows the residual consistently tracking the collisionless component, modified dynamics has a dark component with a spatial distribution, which is most of what a particle hypothesis claims. Speculative If it tracks the gas, the residual is baryonic and MOND is in much better shape than either 2026 paper suggests.

Established Tier three is already running and is delivering a null with a measurable price per unit of reach. Direct detection at LUX-ZEPLIN and XENONnT continues toward roughly a thousand live days by 2028. Frontier The informative quantity is no longer the limit but its derivative: the exposure needed per decade of cross-section is now rising steeply at low mass, and the experiments say so in print. Speculative The test that would change this is directional: an instrument that resolves recoil direction separates a solar neutrino from a halo particle by where it came from rather than by how much energy it deposited, and no such detector exists at the required target mass.

Speculative Tier four is the external field effect, the one prediction MOND does not share with any dark-matter model. A dwarf galaxy’s internal velocity dispersion should depend on the external field of its host, so two otherwise identical dwarfs at different galactocentric distances should differ systematically. Frontier Claimed detections exist in rotation-curve declines and dwarf dispersions and sit at the edge of the systematics. Handwave A pre-registered survey of matched dwarf pairs is the version of this test that would count, and it would be a genuine falsification test rather than another fit.

Frontier Tier five is a theory experiment with no instrument: publish, in advance, what a relativistic MOND predicts for the growth rate and the lensing amplitude. Fitting the microwave background after the fact costs a theory nothing. Speculative A pre-registered prediction for a quantity not yet measured to the required precision is cheap, and its absence is the strongest available evidence that the relativistic constructions are being fitted rather than tested.

Established Tier six has already returned and is the cleanest negative result in the subject. The νHDM simulation programme took a MOND cosmology seriously enough to run it at 800 h⁻¹ Mpc and found bulk flows excluding it at more than 5σ, with a best-fit Hubble constant near 56 km s⁻¹ Mpc⁻¹. Speculative The lesson generalises past MOND: the authors conclude that hot dark matter is unlikely to work regardless of the gravity law, which closes the most developed hybrid position in the field.

Speculative Tier seven is the dynamical-friction test at dwarf-galaxy scale, repeated with feedback models fixed in advance. The globular-cluster result favours Newton for clusters above 4 × 10⁵ solar masses and admits dependence on the supernova prescription. Frontier Running the same comparison with feedback parameters registered before the MOND and Newtonian versions are compared would convert a suggestive result into a test.

Handwave Tier eight is the experiment nobody can design: a positive detection. Everything above is a way of narrowing, and the subject has produced no candidate signal that survived replication in forty years. Speculative A confirmed particle detection ends this brief; a confirmed wide-binary anomaly, confirmed external field effect and a baryonic account of the cluster residual together would end it the other way. Handwave Neither outcome is scheduled.

Established Almost nothing in this brief requires a new instrument; most of it requires analysis of data that exists. The radial acceleration relation came from an infrared photometric survey and archival radio rotation curves; the wide-binary dispute is fought entirely on a public astrometric catalogue; the cluster work uses lensing reconstructions from telescopes built for other purposes. That is unusual and it is why the field’s bottleneck is methodological.

Frontier The exception is direct detection, which is an industrial programme. Ten tonnes of liquid xenon at part-per-quadrillion radiopurity, in a deep underground hall, with photosensors whose own radioactivity is a design constraint, running for 1,000 live days. The next generation is a factor of three to five larger, and the neutrino measurement described in section 3 means its incremental reach at low mass is already known to be poor.

Speculative A directional detector is the one piece of hardware that would change the terms. Recoil-direction sensitivity requires either a low-pressure gas time projection chamber, whose target mass scales badly, or a solid-state track-reading scheme that does not exist at scale. Handwave Nobody has demonstrated either at the tonne scale, and until somebody does, the statement that the neutrino background can be beaten is an engineering assertion rather than a result.

Frontier The astrometric requirement is set by Gaia and its successors. Detecting the tertiary companions that contaminate wide-binary samples needs astrometric wobble solutions rather than statistical modelling, and that is a mission-lifetime question. The instrument side of the survey programme is treated in Mega-Telescopes, and the quantum-limited metrology that underpins the laboratory end appears in Precision Quantum Sensing.

Established This brief and Hidden-Sector Searches divide the same problem at a seam. That brief is the atlas of excluded particle parameter space — axion haloscopes, beam dumps, the scan-rate scaling that governs how fast a magnet can cover frequency. It asks where the particle is not. This one asks whether the search has a target, and its decisive test involves no particle physics at all.

Frontier Neutrino physics has become the limiting factor rather than a neighbour. The solar boron-8 flux is now a background in every xenon detector, which means dark-matter instruments have become neutrino observatories whether or not they intended to; Neutrino Astronomy treats those detectors as instruments, and this brief treats their backgrounds as a constraint.

Frontier Black-hole astrophysics supplies constraints this field cannot generate. The merger population measured by gravitational-wave detectors bounds primordial black holes over part of their mass range, and the assumptions behind those bounds are treated in Black Hole Physics Applications.

Speculative Quantum gravity is adjacent in motivation and not in method. A theory that modifies gravity at low acceleration and a theory that modifies it at high curvature are answering different questions with the same word, and Quantum Gravity shares almost no evidence with this page. Frontier The genuine overlap is sociological: both are fields where a well-posed theory has gone decades without a discriminating measurement, a pattern Objective Collapse Models documents in a third subject.

Established Collider physics is complementary and is not a substitute. A collider produces candidate particles directly and is blind to the astrophysical question of whether a halo exists; Post-LHC Colliders covers what the energy frontier can and cannot say about this problem.

Established The two programmes are funded on incomparable scales and evaluated by incomparable standards. A tonne-scale direct-detection experiment is a national-facility commitment with a decade of lead time and hundreds of authors; the entire published modified-dynamics literature is produced by perhaps a few dozen active researchers using public archives. Frontier This is not a conspiracy and it is a structural problem: the side with more resources produces more careful predictions, so its failures look like refinements and the other side’s look like refutations.

Frontier The field has no pre-registration culture and it needs one more than most. Two of the three live disputes in this brief — whether the acceleration scale is universal, and whether wide binaries show an anomaly — are disputes about analysis choices on shared public data, which is the exact configuration in which pre-registration works. Speculative The cost is a few months of negotiation and the benefit is that one of the two results stops being publishable, which is also why it has not happened.

Frontier Archival data policy is doing more for this subject than any funding decision. Gaia, SPARC and the public lensing archives are the reason a two-person group can contest a result from a large collaboration. Speculative The corresponding hazard is that the same dataset is re-analysed until somebody finds something, and with no registry of analyses the field cannot count how many times a null was found before a signal was.

Speculative Nobody is institutionally responsible for adjudicating between the programmes. Review panels sit inside one or the other; conferences are separated; the principal public test table comparing the two paradigms is maintained by a partisan, which is better than nothing and is not an institution. Handwave A standing, jointly staffed comparison of predictions made in advance would be the cheapest institutional intervention available to this field, and there is no body whose remit covers it.

Established Nothing in this subject harms anyone directly, which makes its ethical questions entirely about the conduct of inquiry. There is no dual-use pathway, no environmental footprint beyond the ordinary cost of a large detector, and no near-term application.

Frontier The allocation question is real and is rarely stated as an ethical one. Direct detection has absorbed a substantial share of the world’s underground physics capacity for three decades on the strength of a hypothesis that has not been confirmed, while the cheapest decisive test in this brief — a jointly registered re-analysis of an existing catalogue — is unfunded because it belongs to nobody. Speculative A funding system that can commission a 10-tonne cryostat and cannot commission an agreed protocol is allocating by capital intensity rather than by information gained.

Frontier Public communication in this field systematically overstates both sides. The Bullet Cluster is routinely described as proof that dark matter exists, when the same system requires a residual missing mass under either interpretation; and MOND is routinely described as refuted by the microwave background, when what failed was a specific baryons-only model from 1999. Speculative Both simplifications are useful to the people making them, which is why they persist.

Speculative There is a career-risk asymmetry that shapes what gets published. A young researcher who finds against the standard model in this subject is entering a dispute with a large and well-organised counterparty; one who finds for it publishes a confirmation. Handwave This brief cannot measure the resulting publication bias and will not pretend the effect is zero.

Speculative Finally, the honest treatment of a forty-year null is itself an ethical question. The field has not misrepresented its limits — the exclusion curves are published, the neutrino background is announced in the abstract — and that record is a credit to it. The pressure point is what happens when the next generation returns another null, and whether the community says so plainly.

Speculative If modified dynamics were right, the practical consequences for a civilisation would be close to nil and the conceptual ones would be large. No engineering follows from a change to the force law at 10⁻¹⁰ m s⁻²; that acceleration is twelve orders of magnitude below anything a machine encounters. What changes is the inventory: about a quarter of the mass-energy census would be deleted and the universe would be made of what we can see.

Established If the particle is right, the consequence is also modest and it is already priced in. A halo of weakly interacting particles does not become a resource; the local density is around 0.4 GeV per cubic centimetre, which is a few hydrogen atoms’ worth of mass per litre and interacts too feebly to be collected. Handwave Any claim that dark matter becomes an energy source is arithmetic nobody has done.

Frontier The real civilisational content of this subject is what it demonstrates about how a science handles a long null. Forty years, several billion dollars, an instrument three million times more sensitive than its ancestors, and no signal — and the field has responded by publishing the limits rather than by quietly changing the subject. Speculative That is a working epistemic culture, and it is more portable than any result this brief could report.

Speculative There is also a cautionary reading, and it belongs here rather than in section 11. A paradigm that can absorb every null by moving a parameter, and an alternative that can absorb every failure by adding a component, together describe a subject that could continue indefinitely without resolving. Handwave The aether survived for decades on exactly this structure and ended when an experiment was designed that no version of it could pass; this subject has not yet designed that experiment, and section 7 is an argument about which candidate comes closest.

These horizons track what could be measured and by whom, not what could be built, because nothing here gets built:

• 10 yr: Frontier Expect the wide-binary dispute to be settled or abandoned on Gaia DR4 astrometry, which will identify a large share of contaminating triples directly rather than statistically; expect it to be settled badly if no joint protocol is agreed first. Frontier Expect direct detection to reach roughly a thousand live days and to report a deeper null with visibly diminishing returns below 10 GeV/c². Speculative Expect the cluster-residual question to move from one famous merger to a sample, and to be the place where modified dynamics is most likely to concede a dark component.
• 25 yr: Speculative This is the window in which the particle programme either detects something or exhausts the accessible weakly-interacting parameter space and hands the question to axion and ultralight searches, which belong to Hidden-Sector Searches rather than here. Speculative A relativistic modified-gravity theory that made a pre-registered prediction for the growth rate and survived it would be the first genuine theoretical advance this subject has had since 1983. Handwave The modal outcome is neither: two deeper nulls and a literature that has grown without converging.
• 50 yr: Handwave On this horizon the plausible development is consolidation rather than discovery — either the halo is detected and this brief becomes a historical note, or the field accepts that the acceleration scale is a fact requiring explanation and stops treating the two positions as rivals. Speculative A third possibility, and the one this brief takes most seriously, is that the answer turns out to be both: a particle that clusters and a regularity in galaxy formation that produces the scale, with neither side having been wrong about its own evidence.
• 100 / 250+ yr: Handwave Beyond useful forecasting. The defensible statement is structural: a hypothesis bounded from below by exclusion curves and from above by consistency requirements either gets measured or gets abandoned, and this one has been bounded from both sides for about fifteen years. Handwave Any specific claim about what a confirmed dark-matter detection would be used for is an assertion, and this page does not make one.

• Depends on One typed dependency on this map and it is not the obvious one. Hidden-Sector Searches supplies the exclusion atlas this brief treats as an input rather than a rival result — without it there is no measure of how much particle parameter space has actually been deleted, and the case for alternatives rests on that number. Neutrino Astronomy supplies the solar flux measurement that is now the limiting background, and Multi-Messenger Astronomy supplied the gravitational-wave speed constraint that eliminated a family of relativistic modified-gravity theories. None of the three blocks progress here; all three set the terms.
• Requires (not on this map) Six conditions, in the order of the tokens above. The protocol. An agreed, registered specification of acceleration bins, quality cuts, triple-system model and calibration sub-sample, eccentricity prior and reported statistic, fixed before Gaia DR4 is unblinded; without it the same catalogue will again yield more than 5σ in both directions. The relativistic theory. A modified-gravity construction that states, before the measurement, what it predicts for the third acoustic peak, the growth rate and the lensing amplitude — fitting these afterwards costs a theory nothing and is what the current constructions do. The residual. A baryonic or otherwise non-particulate account of the 3.4 × 10¹⁴ solar masses that MOND still misses in one cluster, which must be collisionless and galaxy-centred to match the lensing. The directional detector. A tonne-scale instrument that resolves nuclear-recoil direction, the only demonstrated route past a background now measured at 3.3 and 4.5 standard deviations; no such device exists at that mass. The target material. Ten to fifty tonnes of radiopure xenon and argon, a supply constraint rather than a physics one, on decade lead times. The funding line. An evaluation regime in which a published null and a registered re-analysis count as delivered results, which is the cheapest item on this list and the one whose absence does the most damage.
• Enables A confirmed detection would convert Hidden-Sector Searches from a mapping exercise into a measurement programme and would give Post-LHC Colliders a mass scale to aim at. A confirmed modification of dynamics would enable nothing engineering-wise at all — the scale is twelve orders of magnitude below any machine — and would force a rewrite of every structure-formation calculation. No typed enabling edge is claimed, because neither outcome is scheduled.
• Adjacent Within this map: Black Hole Physics Applications for the primordial-black-hole constraints, Quantum Gravity for the other long-running theory without a discriminating measurement, Objective Collapse Models for the same pattern in a third subject, Mega-Telescopes and Precision Quantum Sensing for instruments, Quantum Materials for low-background fabrication, and Cosmological Tensions for the expansion-rate measurements any MOND cosmology must also reproduce, Philosophy of Science and Scientific Revolutions for the question of what counts as an explanation, which is half this dispute.

Established “The Bullet Cluster proved dark matter exists and killed MOND.” Half right, and the half that is wrong is the half people repeat. The lensing offset is real, measured at about 8σ, and it is genuinely hard for any force law that tracks the baryons, because the plasma holds most of the baryonic mass and the potential does not peak on it. Frontier But the same system requires unseen mass under both interpretations: MOND misses a factor of two to three in clusters generally, and a 2026 analysis of this cluster puts the residual at 3.4 × 10¹⁴ solar masses. Frontier A second 2026 paper argues the offset itself is reproduced in QUMOND because compact galaxies generate 48% of the phantom mass while holding 7% of the baryons. Speculative The accurate summary is that the Bullet Cluster is a serious problem for pure MOND and not a proof of a particle, and that MOND’s own literature is divided about which part of it hurts.

Established “The cosmic microwave background rules out MOND.” What it ruled out was a specific model, and the distinction matters. The baryons-only no-CDM model predicted the first-to-second peak ratio correctly and in advance and then failed the second-to-third ratio badly — its own author’s word is “way off”. Speculative That falsifies a 1999 model, not the idea of modified dynamics; relativistic constructions published since fit the spectrum. Handwave Whether they do so by physics or by introducing a field that behaves like dark matter at recombination is the unresolved question, and both camps have an interest in not asking it.

Established “Dark matter has been detected.” It has not. Every reported direct-detection excess has either failed replication or been identified as a background. What has been detected in xenon detectors, at 4.5 and 3.3 standard deviations, is coherent scattering of solar neutrinos — a Standard Model process, and the background that now limits the search.

Frontier “The radial acceleration relation proves a fundamental acceleration scale.” This is the claim this brief was commissioned to check, and it is contested in the refereed literature in both directions. One analysis of 175 galaxies finds no credible variation in the scale; another of 193 galaxies from overlapping data rejects universality at more than 10σ. Speculative Both cannot be right, the disagreement is about priors and sample cuts rather than observations, and a reader told only the first has been told half of a live dispute.

Speculative “The neutrino fog is a hard floor and direct detection is over.” The published statements are weaker than that. What the collaborations report is diminishing returns — 93% more exposure for 10% more reach at 5 GeV/c² — not impossibility, because directional and temporal information distinguishes a solar neutrino from a halo particle in ways that energy alone does not. Handwave This brief read no source that settles whether that information can be collected at the required target mass, and so does not claim the fog is either passable or final.

Frontier “Wide binaries have settled it.” They have produced more than 5σ in both directions from the same catalogue. One group reports a velocity boost of 1.3 to 1.6 below 10⁻⁹ m s⁻²; others report consistency with Newton. Speculative The entire difference is in quality cuts and the modelling of undetected triples, and each side has published a specific technical objection to the other’s procedure. Handwave A result that flips sign under a defensible change of cut is not yet a measurement.

Speculative “MOND has no relativistic version, so it is not a theory.” This was a fair objection until 2004 and is not one now. Relativistic completions exist and one of them is cited by MOND’s leading advocate as reproducing the microwave background. Frontier The 2017 measurement that gravitational waves and light travel at the same speed to about a part in 10¹⁵ did eliminate a family of these theories, which is a real and underappreciated result. Handwave The surviving objection is not that no relativistic theory exists but that the ones that do have enough freedom to fit what they are shown.

Frontier “The small-scale problems have been solved by feedback.” Some have and the important one has not. Supernova feedback produces dark-matter cores above a stellar-to-halo mass ratio of about 10⁻⁴ and does very little below it, so the faintest dwarfs remain unexplained; and the diversity of rotation-curve shapes at fixed circular velocity, coexisting with a relation as tight as 0.057 dex, is the problem that has not yielded.

Handwave “One of these sides is obviously right and the other is cranks.” The strongest thing this research found is that both positions are in retreat. The particle programme has spent three decades and a factor of three million in sensitivity to arrive at an irreducible neutrino background with nothing found; the modified-dynamics programme has arrived at a model whose best-fit Hubble constant is 56 km s⁻¹ Mpc⁻¹ and which is excluded at more than 5σ by bulk flows. Speculative A reader who leaves this page believing either side has closed has read it wrong.