Weighed, Not Named

There is a way of weighing the universe that never asks what it is weighing. Record how fast the galaxies in a cluster are moving and their speeds constrain the mass holding them together: a bound system tolerates only so much motion, and anything faster leaves. The number that comes back has units of mass and no other content. Weighing and identifying are separate operations, and for ninety years only one of them has worked.

The first to run that arithmetic and get a strange answer was Fritz Zwicky in 1933. Applying the virial theorem to the Coma Cluster — a thousand-odd galaxies, eight of them with measured Doppler shifts — he found it needed hundreds of times more mass than its light accounted for. He named the difference — dunkle Materie, borrowing a phrase astronomers already used — and the field left it alone for decades. That was not obtuseness. A discrepancy in a total says nothing about where it came from: the virial theorem assumes a settled, bound system, and the sample was eight galaxies out of a thousand. A gap that large is as easily a fault in the method as a fact about the sky, and naming it does not tell anyone which.

What made it a fact about the sky was repetition at another scale. Vera Rubin and W. Kent Ford mapped 67 regions of the Andromeda Galaxy in 1970, expecting orbital speeds to fall off with distance the way the planets' do around the Sun. They stayed flat. By 1980 Rubin, Ford, and Norbert Thonnard had the same result across 21 spiral galaxies, from the small NGC 4605 to the vast UGC 2885. The gap had not grown; it had become ordinary. One anomalous cluster is a candidate for error. A property shared by every galaxy anyone measures is a property of galaxies.

Everything in that record, though, is an inference from gravity, and gravity is a theory rather than an observation. In 1983 Mordehai Milgrom proposed the other available reading: no unseen substance, only a change in how gravity behaves at the very low accelerations of a galaxy's outskirts. Modified Newtonian Dynamics reproduces flat rotation curves from visible mass alone, elegantly, though it has never accounted as well for the motions inside galaxy clusters. Two accounts of one set of books — a missing entry, or an error in the arithmetic — and for two decades the curves alone could not decide between them.

The separation came from a collision. Douglas Clowe's team studied the Bullet Cluster in 2006, where two clusters had passed through one another, and used gravitational lensing to locate the mass. Most of it had gone straight through, leaving behind the hot gas that carries most of the ordinary matter — a split of eight standard deviations between the mass and the visible material. Modified arithmetic cannot produce that: if visible matter is all there is, gravity has to stay where the matter stayed. The offset is the accounting made spatial: the missing entry has a location.

So the first question is settled: what is absent is a substance, and by now a well-described one. It carries roughly 85 percent of the universe's mass, sits at a density near 26 percent of the total, and the microwave background that Planck mapped confirms the same ΛCDM picture independently at cosmological scales. Its distribution has been traced across billions of light-years. What is missing is not a quantity or a map. It is a noun.

The instrument that would supply one has been running for decades. The leading candidate, the Weakly Interacting Massive Particle, is attractive because a particle of roughly nuclear mass, between 10 and 1000 GeV, would have frozen out of the early universe close to the observed abundance. Detectors were built to wait for a nudge, not a glow: ultra-pure liquid xenon or germanium under kilometers of rock. Nothing has arrived. Economy is a reason to look somewhere first, never evidence that anything is there.

That is a result, not a failure. Each null run at greater sensitivity narrows where the particle can be, which is most of what a search accomplishes. But the technique has a stopping point built in. As detectors grow more sensitive they approach the neutrino floor, where ordinary background neutrinos produce recoils very nearly indistinguishable from the one being hunted. That is not a limit of care or funding. It is the noise the universe itself supplies, and it does not come down. Detectors that read the direction a nucleus recoiled could probe below it; those now running hold a few hundred grams.

Meanwhile the candidate list has widened — the axion, proposed for an unrelated problem in particle physics; dark sector models in which the missing mass is a family, not a species. A widening list of candidates against a narrowing searchable volume is what a field looks like when its best-motivated guess declines to appear.

None of which is a shortage of effort. Four centuries of astronomy were built around gathering light, and this gives none off, so every inherited instrument is the wrong shape for the question. A mass, a fraction, a map, a role in cosmic history — almost everything a ledger entry requires, except the identity of the thing entered.