Ancient rocks from Western Australia suggest Earth was already dragging surface water into its mantle and using it to generate volcanoes 3.1 billion years ago — through a primitive recycling process that may have preceded modern plate tectonics

Pillow lavas look like a heap of stone beanbags, slumped one over another where molten rock hit cold seawater and stiffened on the spot. The ones exposed across a stretch of Western Australia’s Pilbara have been sitting there, largely undisturbed, since long before anything on this planet had a leaf or a nucleus.

Rock that old is normally a lost cause. Three billion years of heat and pressure scramble a lava’s original chemistry until reading it becomes guesswork. The Pilbara Craton mostly dodged that fate, and a study published in Nature Communications describes it as the least altered surviving crust older than 2.8 billion years anywhere on Earth. Cooling joints, gas bubbles and pillow outlines are still visible in outcrop, which for rock of that age is close to miraculous.

So an international team led by Adelaide University geochemist Eric Vandenburg went out with hammers and sampled a ten-kilometre-thick slice of it.

What the hammers turned up

The target was the Whundo Group, a stack of volcanic rock laid down between 3.13 and 3.10 billion years ago. Roughly 30 million years of eruptions, preserved in sequence like pages in a book nobody had read closely. Sampling it at tight spacing gave the team three distinct families of lava: tholeiites, calc-alkaline basalts, and boninites.

That third one is the interesting one. Boninite is a rare, water-rich, magnesium-heavy lava named after the Bonin Islands in the western Pacific, and today it erupts almost exclusively where one tectonic plate is grinding down beneath another. The Whundo examples are the oldest extensive boninites known.

The number that changed the story

Water does something specific to mantle rock. It drops the melting point, roughly the way a scatter of salt turns road ice to slush on a day that should be far too cold for slush. Writing in The Conversation, the researchers used exactly that comparison to explain why volcanoes cluster along subduction zones and not much of anywhere else.

Work backwards from the chemistry and you can estimate how wet the source rock must have been. The team’s calculations put the mantle that fed the Whundo boninites at somewhere between 0.8 and 1.5 percent water by weight. Primitive mantle sits near 0.11 percent. Depleted mantle is closer to 0.01. Mantle rock beneath a present-day arc volcano holds between 0.1 and 2 percent, which lands the ancient figure squarely inside today’s range, and the paper stresses that its estimates are minimums.

Water in that quantity does not turn up in the mantle by accident.

Subduction without the plates

So how do you get seawater that deep without plate tectonics?

Conventional subduction almost certainly was not available yet. A hotter young Earth meant weaker, more ductile crust, and geodynamic models have long held that slabs sliding cleanly beneath one another would have been mechanically very hard to sustain that far back.

The alternative goes by the name dripduction, developed across several modelling studies including work by Oliver Nebel and colleagues in Precambrian Research. Picture a section of waterlogged crust growing dense enough to sag, then founder, sinking into the hotter mantle below as a lopsided blob rather than a rigid descending sheet. No continuous plate boundary, no conveyor belt, nothing built to last. Gravity doing locally, in short bursts, what plate motion does globally today.

Messy as that sounds, the chemical result at Whundo looks remarkably like the real thing. Fluids released by the foundering crust supplied as much as 93 percent of certain trace elements in the mantle source that later melted to make the boninites, per the team’s mass-balance calculations. The pattern repeats all the way up the sequence, which means whatever was happening kept happening for tens of millions of years.

One patch of rock, one paper

One study is one study, and this team says so plainly. The work covers a single volcanic package of about 114 square kilometres. Nobody involved claims that modern plate tectonics switched on in the Pilbara 3.1 billion years ago, and the paper opens by acknowledging that arc-like volcanism from the Archean eon, the stretch running from roughly four to 2.5 billion years ago, remains disputed evidence for subduction.

There is give in the modelling, too. The study notes outright that its preferred explanation is not the only one that fits, and offers a second scenario reproducing the same boninite chemistry by a different route. Full datasets sit on Zenodo for anyone inclined to try a third.

What the evidence supports is narrower and still substantial: at one well-preserved location, surface water was reaching melting depths and building arc-style volcanoes well before the machinery normally credited with that job existed.

The crust that isn’t there

Between 50 and 90 percent of the evolved crust that models say the Archean should have produced is missing from the record. Cratons older than 2.8 billion years, the ancient cores of continents, account for about seven percent of the continents today.

The Adelaide University announcement framed the finding as evidence that key planetary processes were already running early. Vandenburg and colleagues push further, suggesting the crust dripduction built was thin and chemically primed for destruction, which is precisely why so little of it survived.

Follow that through and the rock record starts looking like a survivor bias problem. What is still here is what was too thick, too buoyant or too lucky to get swallowed. Whundo may be less a discovery than a rare witness, and the process it records might have been the ordinary state of affairs on a planet that ate nearly all of its own evidence.