What happened in the first three minutes?

In the first second the universe was hot enough that protons and neutrons converted into each other freely. As it cooled, the conversions stopped and the ratio froze at roughly one neutron to six protons, set by the small mass difference between them and the temperature at which that happened. Free neutrons then decayed for the minutes it took fusion to start, and by the time it did the ratio was about one to seven.

Fusion could not begin immediately. Any deuterium formed was broken apart by the ambient radiation almost at once, a delay known as the deuterium bottleneck. Only when the temperature fell far enough did deuterium survive long enough to be built into helium.

By the time the universe was a few minutes old it was too cool and too thin to fuse anything else, and the composition was fixed until stars began to change it.

The calculation has essentially one free parameter: how much ordinary matter there was, expressed as the ratio of baryons to photons. Fix that and the yields of all the light nuclei follow from nuclear physics measured in laboratories.

The same parameter is measured independently from the microwave background, which is a snapshot taken hundreds of thousands of years later by a completely different method. The two agree, within a margin that turns on laboratory rates for the reactions that burn deuterium. One team (Yeh, Olive and Fields) finds them consistent; another (Pitrou and colleagues) reports a mild tension of about 1.8 standard deviations.

That agreement is one of the strongest pieces of evidence that the early universe was hot and dense in the way the model says, because there is no reason a nuclear calculation and a pattern of temperature ripples should return the same number by accident.

The predicted amount of lithium-7 is about three times what is measured in the atmospheres of metal-poor stars in the halo of the galaxy. The Particle Data Group puts the gap at a factor of 3.3, a 4.4 sigma discrepancy, and it has survived decades of improved measurements on both sides.

One ordinary explanation has been all but closed off. The reviews work through the reactions in which an unmeasured resonance could have destroyed more of the beryllium-7 that later becomes lithium, and the experiments that looked have found none. That leaves errors in the stellar measurements, something those stars do to their own lithium, or physics the calculation does not contain. The Particle Data Group now counts the stars as the leading explanation, and still calls the problem unresolved.

Because the yields depend on how fast the universe expanded, they constrain anything that would have added to the energy density at the time, including additional species of neutrino.

It is a constraint rather than a detection. Counting neutrino families from the light elements returns close to the three the Standard Model has: 2.898 plus or minus 0.141 at 68 per cent confidence.