Diamond-anvil cells are the standard tool for studying matter at extreme pressures. The sample sits between two diamonds. The diamonds are not thought to participate in the physics — they're the container, not the experiment. For high-pressure ice, this assumption is wrong.
Large-scale molecular dynamics simulations with neural network potentials now show that the diamond-ice interface significantly lowers the superionic transition temperature. In superionic ice, the oxygen atoms form a lattice while hydrogen atoms flow like a liquid through it — a state relevant to the interiors of Uranus and Neptune. The interface doesn't just contain this transition. It catalyzes it.
More dramatically, the interface induces a spontaneous structural transition from body-centered cubic (bcc) to face-centered cubic (fcc) ice via the inverse Bain mechanism — a transformation the bulk sample wouldn't undergo at the same pressure. The diamond doesn't observe. It acts. The container reshapes the phase diagram of its contents.
This means that decades of high-pressure ice experiments may carry a systematic bias: the measured transition pressures and temperatures reflect the system-with-interface, not the bulk material alone. The discrepancies between theoretical predictions and experimental measurements of ice phases aren't necessarily errors in theory. They may be accurate measurements of the wrong system — the ice-diamond composite rather than pure ice. The instrument changed the answer.