The standard model of Earth's interior was simple about water. The upper mantle holds some — olivine transitions to wadsleyite and ringwoodite at depth, and these minerals can incorporate hydroxyl groups into their crystal structure. The transition zone (410-660 km) is relatively wet. Below that, bridgmanite dominates the lower mantle from 660 km to the core-mantle boundary at 2,900 km. And bridgmanite, according to decades of high-pressure experiments, holds negligible water.
The conclusion seemed settled: the lower mantle is dry. Earth's largest geological compartment — over half the planet's volume — was written off as a water reservoir.
Lu, Du, and colleagues (Science, 2025) found the experiments were wrong. Not because the measurements were imprecise, but because the temperatures were too low.
Bridgmanite forms during Earth's magma ocean phase, when the planet's interior is at roughly 4,000-4,100 degrees Celsius. Prior high-pressure experiments ran at lower temperatures because achieving extreme heat and extreme pressure simultaneously is technically brutal. The team at Guangzhou Institute of Geochemistry built a diamond anvil cell with laser heating and high-temperature imaging that pushed to approximately 4,100 degrees. At those conditions — the actual conditions under which bridgmanite first crystallized — the mineral's water storage capacity increased dramatically. Five to one hundred times greater than earlier estimates.
The relationship between temperature and water retention is not linear. Bridgmanite's ability to incorporate hydroxyl groups into its lattice jumps sharply above a threshold. Below that threshold, at the temperatures prior experiments reached, water storage looks negligible. Above it, at the temperatures that actually existed during Earth's formation, bridgmanite becomes a sponge.
Running these revised storage numbers through a magma ocean crystallization model, the lower mantle could retain 0.08 to 1 ocean's worth of water. One ocean, hidden in a mineral that was classified as dry. The lower mantle may be Earth's largest solid-state water reservoir — not despite being mostly bridgmanite, but because of it.
The detection required instruments that didn't exist during the original experiments: cryogenic three-dimensional electron diffraction, NanoSIMS, atom probe tomography. The water is incorporated at the nanoscale, distributed through the crystal lattice. You can't see it with techniques that average over larger volumes.
What fascinates me is the error structure. The original experiments were correct at their own temperatures. Bridgmanite at 2,000 degrees stores negligible water. This is true. But the extrapolation — “therefore bridgmanite in the deep Earth stores negligible water” — requires the assumption that the laboratory conditions are relevant to the geological conditions. They weren't. The temperature gap between the experiment and the reality was the gap between dry and wet.
This is a general pattern in high-pressure geophysics. You study materials under the most extreme conditions you can achieve in the lab, and then you generalize to conditions that are more extreme still. Usually this works. Phase boundaries shift smoothly. Properties extrapolate. But sometimes there's a nonlinearity — a threshold, a phase transition, a sharp change in capacity — that lives in the gap between what you can measure and what you need to know. The gap is invisible until someone closes it.
Du's team closed it with better engineering, not better theory. The physics was always there. The mineral was always capable. The planet had been storing water in its largest compartment for 4.5 billion years. We just hadn't heated the experiment enough to see it.
Lu, Yang, Long, Xian, Li, and Du, "Substantial water retained early in Earth's deep mantle," Science (2025). DOI: 10.1126/science.adx5883. Guangzhou Institute of Geochemistry, CAS.