What is most of the universe made of?

Vera Rubin and Kent Ford measured how fast gas orbits in spiral galaxies, first in Andromeda and then, with Norbert Thonnard, in twenty-one more, and found that the speed does not fall away towards the edge: in most of them it was still rising slowly at the farthest point measured. If the visible mass were all the mass, the outer gas would be moving too fast to stay bound. Either there is unseen matter, or gravity behaves differently at those scales.

Gravitational lensing, the growth of large-scale structure, and the pattern of fluctuations in the cosmic microwave background all independently require substantial non-luminous mass. The Planck satellite’s measurements give roughly 5% ordinary matter, 26% dark matter and 68% dark energy.

In the Bullet Cluster, two galaxy clusters collided. Their hot gas, which glows in X-rays and holds most of their ordinary matter, was slowed by the collision; their galaxies passed straight through. Gravitational lensing shows the mass travelling with the galaxies, not with the gas. The mass and most of the ordinary matter are in demonstrably different places.

Observations of distant supernovae in the late 1990s showed that cosmic expansion is speeding up rather than slowing under gravity. The cause is labelled dark energy. Naming it did not explain it.

The leading candidate class (weakly interacting massive particles) has been searched for with increasingly sensitive underground detectors for decades, with no confirmed detection. Each null result narrows the allowed parameter space without ruling out the idea.

Axions, sterile neutrinos and primordial black holes remain alternatives, the last two now tightly constrained, and each is being searched for with different instruments.

Modified Newtonian dynamics and its relativistic descendants reproduce galaxy rotation curves impressively well with no dark matter at all. They struggle with galaxy clusters, where even their proponents find mass still missing. Most versions also fail on the cosmic microwave background, though the authors of one relativistic theory, published in 2021, report that theirs fits it. The mainstream view is that dark matter is real; the alternative is not dead.

Measurements of the expansion rate from the early universe and from nearby distance indicators disagree by more than their stated uncertainties: by five standard deviations, according to the SH0ES team, which has tested its own analysis in nearly seventy variants. A rival team calibrating the nearby route with red giant stars gets a value between the two and sees no need for new physics. The gap may be a systematic error nobody has found, or a sign that the standard cosmological model is incomplete.

No particle in the Standard Model can account for it. No detector has recorded it beyond dispute: one experiment reports a seasonal signal, and two others built from the same kind of crystal to test it do not see it. Its existence is inferred entirely from its gravitational effects, which are measured precisely and consistently across many independent methods.

The simplest description is a constant energy density of empty space. Calculating that value from quantum field theory gives an answer wrong by many tens of orders of magnitude. Steven Weinberg’s 1989 review of the problem put the gap at some 120.

If dark matter is a particle that touches ordinary matter through something more than gravity, as WIMPs would, the next generation of detectors will be ten to a hundred times more sensitive than today’s, and is meant to keep going until neutrinos arriving from space start to drown out any signal. It may find the particle. If it does not, much of the range where the favoured candidates were expected will be closed, which would itself be a major result.

Dark energy may be the cosmological constant, or it may be a field whose strength changes over time. In 2025 the DESI survey, combining its own measurements with the microwave background and supernovae, reported a preference for the latter at 2.8 to 4.2 standard deviations, depending on the supernova sample. Corrections to the supernova samples since have narrowed that to 3.2 to 3.4; the team behind the sample that gave 4.2 now calls its result a weak preference, and George Efstathiou has argued that a small offset between supernova samples could bring the most discrepant one into line with a constant. These are hints, not conclusions, and the Particle Data Group’s dark energy review treats them with “healthy skepticism”.

Dark matter, dark energy and the expansion-rate discrepancy might be three symptoms of one missing piece of physics rather than three separate problems. No theory of this kind has yet been shown to fit all the data. Even replacing dark matter alone with modified gravity has so far worked on some scales and failed on others.