Earth’s oceans may have been forged from a rain of dry pebbles

Where did Earth’s water come from? For decades, the leading candidates have been extraterrestrial: icy comets slamming into the young planet, or water-rich asteroids delivered from the outer solar system. But a new laboratory study published in Nature Geoscience suggests a radically different possibility, one that requires no delivery from afar at all. According to Susmita Garai, Zachary D. Sharp, Peter L. Olson and Anthony M. Gargano of the University of New Mexico and the Lunar and Planetary Institute, Earth’s oceans could have been manufactured in place, generated chemically inside the scorching hydrogen atmosphere that enveloped our planet while it was still being assembled from millimetre-to-centimetre-sized silicate pebbles.

The idea hinges on a growing consensus in planet-formation theory that terrestrial planets may not have grown primarily through violent collisions between large rocky bodies. Instead, many models now favour pebble accretion, a process in which a growing planetary embryo sweeps up a continuous stream of small, pebble-sized particles drifting through the gas of the protoplanetary disk. As these pebbles rain down onto the young Earth, they pass through a dense primordial atmosphere rich in hydrogen, captured directly from the nebula. That atmosphere, the researchers argue, was not a passive bystander. It was a chemical reactor.

To test this hypothesis, the team did something deceptively simple but technically demanding: they melted dry, anhydrous mafic rocks, including a mid-ocean-ridge basalt composition representing typical silicate planetary material, in streams of hydrogen gas under controlled laboratory conditions. The setup simulates what a molten pebble would experience as it plunged through the hot, hydrogen-rich envelope above a growing terrestrial planet. The temperatures were extreme, and the chemistry was deliberately reducing, meaning the hydrogen gas was poised to strip oxygen from any oxide compounds it encountered.

The results were striking. Within several hours, the experiments produced large amounts of water. The mechanism is a well-understood redox reaction with profound planetary consequences: hydrogen molecules react with iron oxide dissolved in the silicate melt, reducing ferric and ferrous iron to metallic iron while the liberated oxygen combines with hydrogen to form water vapour. In chemical terms, FeO in the silicate plus molecular hydrogen yields metallic iron plus H2O. Every unit of iron oxide reduced is a unit of water born. The measurements showed that the reduction process is efficient, converting a substantial fraction of the iron oxide in the starting material into metal on timescales of hours, which is geologically instantaneous.

This single reaction may solve several long-standing puzzles about Earth’s composition at once. The first concerns water itself. If pebbles settling through a hydrogen atmosphere were repeatedly melted and reduced, the water vapour generated would accumulate in the planet’s envelope and eventually be incorporated into the growing planet or retained after the nebular gas dissipated. The authors calculate that this process could produce oceans’ worth of water, making it a genuinely endogenous source, manufactured from the planet’s own raw ingredients rather than imported from beyond the frost line. That would sidestep the timing problems that plague delivery models, in which water must arrive late enough to survive but early enough to participate in Earth’s subsequent evolution.

The second puzzle involves phosphorus. Earth’s mantle is strikingly poor in phosphorus compared with the mantle of Mars, and geochemists have struggled to explain why two neighbouring planets, built from broadly similar solar-system materials, should differ so much in this biologically essential element. The experiments revealed that the metallic iron produced by hydrogen reduction strongly sequesters phosphorus, pulling it out of the silicate melt and into the metal. Because the dense metallic liquid would subsequently sink through the magma ocean to join the growing core, any phosphorus it carried would be efficiently removed from the mantle. Atmospheric reduction during pebble accretion therefore offers a natural mechanism for stripping phosphorus from Earth’s silicate reservoir before the core finished forming.

The third puzzle is the iron-to-magnesium ratio of Earth’s primitive mantle, which is lower than chondritic expectations. Removing iron from the silicate fraction, by reducing FeO to metal and sending that metal to the core, lowers the Fe/Mg ratio of the residual mantle in exactly the direction observed. In this picture, the distinctive chemical fingerprint of Earth’s mantle is not an accident of which building blocks happened to arrive, but the direct consequence of the atmosphere through which those building blocks fell.

The experimental evidence is documented in detail across the study’s figures, which map the conditions under which iron oxide reduction occurs in settling pebbles, track the compositions of the run products against the initial basaltic starting material, and image the quenched melt surfaces with scanning electron microscopy. The reduction efficiency, expressed as the percentage of FeO converted to metallic iron, rises with both time and temperature in the hydrogen streams, and the team quantified the water produced alongside the metal generated. The underlying experimental data and the input data for their pebble-accretion model of Earth are publicly archived in the Dryad repository, allowing other researchers to scrutinise and extend the calculations. The work was supported by the United States National Science Foundation.

The implications extend well beyond our own planet. If Earth-sized worlds can acquire oceans by growing through pebble accretion inside hydrogen-rich envelopes, then the same chemistry should operate on rocky exoplanets forming by the same route, which current models suggest may be the dominant pathway for building planets across the galaxy. Water vapour has already been detected in the atmospheres of hydrogen-rich sub-Neptune exoplanets, and atmospheric reduction of iron oxide provides a physically grounded mechanism linking such observations to surface habitability. A planet does not need to be born wet or bombarded by comets; it needs only dry silicate pebbles, a hydrogen atmosphere, and time. Habitability, in this view, is not a lucky delivery but a step in the construction sequence.

There remain questions to resolve, as with any laboratory simulation of planetary-scale processes. The experiments were conducted at specific temperatures, pressures and hydrogen flow rates, and scaling the results to the full range of conditions in a primordial planetary envelope will require further modelling and measurement. The isotopic composition of Earth’s water, particularly its deuterium-to-hydrogen ratio, will continue to constrain how much water could plausibly have come from nebular hydrogen versus other sources, since water made from nebular hydrogen carries a distinctive isotopic signature. Yet the core result stands on its own: dry rock plus hydrogen gas plus heat equals water, and the young Earth had all three in abundance. The rain of pebbles that built our planet may have carried, invisibly, the recipe for the oceans that made it habitable.

Subject of Research: Experimental simulation of water production by hydrogen reduction of iron oxide during pebble accretion on forming terrestrial planets

Article Title: Experimental simulation of water formation on Earth from dry pebble rain

Article References: Garai, S., Sharp, Z. D., Olson, P. L., & Gargano, A. M. (2026). Experimental simulation of water formation on Earth from dry pebble rain. Nature Geoscience. https://doi.org/10.1038/s41561-026-02118-7

Image Credits: AI Generated

DOI: 10.1038/s41561-026-02118-7

Keywords: water origin, pebble accretion, hydrogen atmosphere, iron oxide reduction, Earth formation, Nature Geoscience, planetary habitability, phosphorus depletion, magma ocean, exoplanets, geochemistry, protoplanetary disk