Vera Rubin and Kent Ford spent the 1970s at the Carnegie Institution of Washington mapping the rotation of dozens of spiral galaxies. The Newtonian prediction was clear: stars far from the galactic centre should orbit slowly — rotational velocity declining as 1/√r in a Keplerian way. The observation was something else. The rotation curves Rubin measured stayed flat far beyond the visible disk, in galaxy after galaxy. Either Newtonian gravity was wrong at galactic scales, or there was enormously more matter in galaxies than was visible — five to ten times more, distributed in a halo. The second interpretation is the mainstream account. Ordinary baryonic matter accounts for only ~5% of the universe's energy budget, and the bulk of the gravitating universe is some form of matter we have never directly detected.
The case for cold dark matter rests on four largely independent lines of evidence. Galaxy rotation curves — Rubin's program — establish that the gravitating mass of any large galaxy extends far beyond its visible disk. Gravitational lensing of background galaxies by foreground clusters is the second line: the lensing mass is several times the visible cluster mass, and the Bullet Cluster — a merging-cluster system where the X-ray-emitting hot gas has been spatially separated from the gravitational-lensing mass — is a particularly clean signature: the lensing mass is not coincident with the gas, which is hard for any modified-gravity theory to explain. Structure formation is the third: cosmological N-body simulations need cold dark matter to match the observed galaxy distribution at all scales; the dominant model is ΛCDM. CMB anisotropies are the fourth: the angular power spectrum of cosmic-microwave-background temperature fluctuations measured by Planck is consistent with ΛCDM and inconsistent with baryon-only universes. Modified-gravity alternatives — MOND, MOG, and successors — explain some of the data well but struggle with the Bullet Cluster signature and the CMB power spectrum. The mainstream view is that dark matter exists, even though we have not yet identified its particle physics. The leading candidates are WIMPs (weakly interacting massive particles, predicted by some supersymmetric extensions of the Standard Model), axions (very-light pseudoscalar particles motivated independently by the strong-CP problem in QCD), and primordial black holes. Direct-detection experiments — LZ, XENONnT, PandaX — have not found WIMPs over the canonical mass range despite forty years of increasingly sensitive searches. ADMX searches the axion mass range using cavity haloscopes; microlensing surveys constrain primordial black holes. Indirect-detection programs (Fermi-LAT gamma-ray searches, IceCube neutrino searches) have placed strong constraints. The gap between evidence-of-dark-matter (overwhelming) and identification-of-dark-matter (zero) is itself a substantial constraint on what dark matter can be.
Two new directions are reshaping the search. The first is axion experiments with new sensitivities — ADMX-G2, HAYSTAC, MADMAX — covering axion mass ranges that were inaccessible a decade ago; an axion detection would simultaneously solve the strong-CP problem and the dark-matter question. The second is gravitational-wave observation of primordial-black-hole mergers as a constraint on that scenario. As of 2024, neither has detected its candidate. Dark matter is the central open problem of contemporary fundamental physics, with the largest single funding line in the field, and four decades of failed searches have begun to change the bets people are willing to place. The next ten to fifteen years should be decisive.