Was the whole planet once frozen?

Two intervals in the Cryogenian, roughly 717 to 660 and 650 to 635 million years ago, left glacial deposits in more than ninety formations on twenty-two palaeocontinents and microcontinents, including rocks that palaeomagnetism places in the tropics. The onset of the first and the end of both are now dated to about a million years, on several of those continents at once.

The glacial deposits are capped, worldwide, by a distinctive layer of carbonate rock whose chemistry indicates a sudden and extreme change in the ocean and atmosphere at the end of the glaciation. How fast the cap itself accumulated is unresolved: its sedimentary structures suggest quickly, and the magnetism recorded in it suggests tens or hundreds of thousands of years.

With the surface frozen, rain stops and the weathering of rock that normally removes carbon dioxide all but stops with it. Volcanoes keep emitting. Over millions of years carbon dioxide builds to something like a hundred times the present level, which is what the geochemistry at the younger termination indicates.

That eventually overwhelms the reflectivity of the ice, and the deglaciation is catastrophic: in a coupled ice-sheet model the ice line runs from the equator to the poles in about two thousand years, into extreme greenhouse warming over an ocean that has just been uncovered, which is what the cap carbonates appear to record.

The loop this describes is what makes the idea testable. It predicts glaciations that were long, that began at low latitude, and that ended everywhere at once under extreme carbon dioxide forcing, which the dating and the geochemistry have increasingly supported since.

The threshold is a model estimate, and the estimates span a factor of ten. One Earth system model starts the thaw in the mid-latitudes instead, with dust darkening the snow, even at present-day carbon dioxide.

A completely frozen ocean is difficult to reconcile with the survival of photosynthetic life, which came through these intervals and diversified afterwards. Refuges have been proposed: thin ice in the tropics, meltwater ponds on the surface, cracks and volcanic areas.

The alternative is that a belt of open water survived near the equator throughout, which preserves the biology at the cost of weakening the escape mechanism. These solutions have to sit next to the runaway without falling into it for millions of years, and still leave the abrupt deglaciation the cap carbonates record. Whether they can depends on the model: in one, a fuller treatment of sea ice leaves no stable belt of open water at all.

The sediments are read both ways. Allen and Etienne argue that glaciers kept delivering sediment to open ocean throughout, so some ocean stayed free of ice. Layers of volcanic ash inside the glacial deposits have been argued to require open water, and argued to have been carried to the tropics by the ice itself.

Candidates include the weathering of a large volcanic province emplaced in the deep tropics, the breakup of a supercontinent bringing rain to its dry interior, and sulphur from that same province loading the stratosphere with reflective aerosol. The Sun was about six per cent fainter than it is now, which is a condition rather than a trigger.

The first two would have drawn carbon dioxide down and moved the climate towards the threshold. The province and the glacial onset are indistinguishable in time, which is suggestive and does not settle which of the mechanisms it supplied did the work.

Animal body fossils appear in the record after the second of these glaciations. Molecular clocks put the lineage earlier, before the first of them, and disputed biomarkers place sponges in the warm interval between the two. On that reading the animals came through the ice rather than after it.

The glaciations are associated with the emergence of animals, and a causal link is not established. The oxygen record between 800 and 400 million years ago is not good enough to defend or refute the idea that these glaciations pushed the atmosphere towards its later state.