Text from the interactive explanation
Ultrahigh pressure phase transition in ice
Israel Osmond, Mikhail A. Kuzovnikov, David A. Lewis, Hannah A. Shuttleworth, Tomás Marqueño, Milo J. Dixon, Cerian E. A. Robertson, Vitan Cherepnalkoski, Bihan Wang, Andreas Hermann, Miriam Peña-Alvarez, Ross T. Howie
Original paperNature Materials · 2026 · Osmond et al.
Past 308 gigapascals, three million atmospheres, the single X-ray ring of cubic ice breaks into three. Turn the screw and watch the crystal shear.
Past 308 GPa the X-ray ring breaks in three.
Distorted ice X · tetragonal
Main X-ray peak: one, asymmetric
Model values from the paper’s equations of state (Suppl. Table 2).
2θ (°) · λ = 0.29106 Å
Phases along the scale: ice VII up to 70 GPa, ice X from 70, distorted ice X from 100, ice XXII from 308 GPa.
1 · Raman spectroscopy · measured
Listen to the oxygen lattice
X-rays mostly see the oxygen atoms; Raman spectroscopy hears how they vibrate. Cubic ice X has exactly one Raman-active vibration, the T2g mode: oxygen atoms moving among stationary hydrogens. The team followed it up to 253 GPa. It broadens, and above 160 GPa it splits into two distinct modes. One allowed mode becoming two means the oxygen lattice is no longer cubic.
89 GPa · one mode (T₂g)
At the same pressures the (110) X-ray peak turns asymmetric. The paper attributes both to a weak tetragonal squash of ice X: c/a ≈ 0.99 at 140 GPa, ≈ 0.97 at 300 GPa. The splitting grows with pressure, Δν ≈ 52 cm⁻¹ at 167 GPa and 69 cm⁻¹ at 253 GPa, and was seen in every run.
Above about 250 GPa, pressure-induced fluorescence swamped the weak signal, so ice XXII’s own Raman spectrum could not be measured. Extrapolated, the two modes line up with the two most intense modes DFT calculates for Pbcm ice.
2 · X-ray diffraction · the deduction in Fig. 3
Name the new ice
Above 308 GPa, within about 10 GPa, the first group of X-ray peaks splits into three. Which crystal structure does that? Put each suspect’s computed pattern on the measured 335 GPa pattern and check it against four clues, as the paper does.
C1 The first peak group splits into three components·
C2 The highest-angle component is the strongest·
C3 A weak extra peak near 10.5° in some patterns·
C4 Nothing at 4.7°·
6 suspects remaining
Pick a suspect to overlay its computed pattern.
The paper’s check: Rietveld refinement of ice XXII at 338 GPa
3 · Equation of state · measured
A small jump
Ice X compresses smoothly, from about 9.4 ų per molecule at 105 GPa to 7.3 ų at 308 GPa. At the transition the volume per molecule drops by 0.04 ų. Small, but a discontinuity in volume is the signature of a first-order phase transition.
Ice X squeezes smoothly from 105 to 308 GPa. Zoom in on the transition to see the step.
Not a strain artefact
Gold powder sat in the same chamber, surrounded by the ice. If uneven stress had split the ice peaks near 308 GPa, the gold reflections would have changed too. Their profiles did not change.
Reversible
When a diamond anvil failed at 338 GPa, the pressure fell to about 39 GPa and the sample went back to ice VII (Supplementary Fig. 7).
Reproduced
Ice XXII was stable to at least 340(5) GPa in the main run. A second run, with poorer data, reproduced the transition and, after correcting its pressure scale, showed ice XXII to at least 355(10) GPa.
4 · The experiment · to scale
Inside the cell
All of this happened in a speck of water a few micrometres thick, squeezed between the tips of two diamonds and probed with a focused synchrotron X-ray beam. Zoom from the anvils to the atoms.
Field of view ≈ 6 mm · two diamond anvils
Culets 15–50 µm across, bevelled to 200 µm; gasket ≈ 6 µm thick; sample chamber ≈ half the culet.
PETRA III, P02.2 (data above 250 GPa): E ≈ 42.4 keV, λ ≈ 0.292 Å, beam focused to 2 × 2 µm.
ESRF, ID15B (below 250 GPa): E ≈ 30.2 keV, λ ≈ 0.410 Å, ≈ 1 µm spot.
Gold grains in the chamber served as the pressure gauge; the cell was rocked ±5° or ±10° during each 60–240 s exposure.
10 runs, 2 crossed into ice XXII. At ~350 GPa the ice’s diffraction signal was only 1/50 to 1/300 of the Compton background from the diamonds.
5 · Theory versus experiment
Theory called it
For almost 30 years, density functional theory (DFT) predicted that ice X must turn into an orthorhombic Pbcm phase. The predicted pressure depends on the exchange–correlation functional, the approximation at the heart of DFT, and varies by almost 50 GPa. Pick one.
Shaded band: measured 308 ± 5 GPa
Select a functional to compare its predicted transition pressure with the measurement.
Next: Pbca near 760 GPa
Ice XXII is predicted to be stable over a vast range. Another orthorhombic form, with Pbca symmetry, is expected around 760 GPa, beyond what conventional static compression can reach.
Hot and superionic
Shock experiments in this pressure range, at temperatures above 2,000 K, find a superionic fcc ice, ice XVIII. Bridging static and dynamic compression to understand the XXII–XVIII transition is an open challenge.
Giant planets
To the authors’ knowledge ice XXII is the densest H₂O polymorph ever observed. Its structure informs models of giant-planet interiors, and the match shows DFT reproduces water under such extreme conditions.
Source & credits
Osmond, I., Kuzovnikov, M. A., Lewis, D. A., Shuttleworth, H. A., Marqueño, T., Dixon, M. J., Robertson, C. E. A., Cherepnalkoski, V., Wang, B., Hermann, A., Peña-Alvarez, M. & Howie, R. T. “Ultrahigh pressure phase transition in ice.” Nature Materials 25, 1850–1854 (2026). doi.org/10.1038/s41563-026-02732-1 · article page
Open Access under a Creative Commons Attribution 4.0 International licence (CC BY 4.0); © The Author(s) 2026. X-ray diffraction data: University of Edinburgh DataShare, doi.org/10.7488/ds/8156.
Measured curves are redrawn from the vector data of Figs. 2a, 2b, 3a, 3c and 4 (digitised; re-coloured, re-scaled and re-laid-out). Crystal coordinates and equations of state from Supplementary Tables 1–2. Crystal renders, computed diffraction patterns and animations are models built from those numbers; the anvil drawing and the ×5 exaggeration are illustrative.
Topics in this iPaper
Physics
through diamond anvil cell experiments combined with synchrotron X-ray diffraction and Raman spectroscopy techniques, we demonstrate that post-symmetrization, the crystal structure of ice X continuously distorts with pressure
Abstract, page 1850
Chemistry
DFT-based computational quantum chemistry has previously predicted that H2O should transform into an orthorhombic structure with Pbcm space group upon compression above 250–350 GPa
Transition to ice XXII section, page 1851
high pressure phase transition
Subjecting ice to pressures in excess of 308(5) GPa, we observe a first-order phase transition to ice XXII
Abstract, page 1850
diamond anvil cell
Pressure was generated using a custom-designed (piston–cylinder type SH) symmetric diamond anvil cell.
Methods, Sample preparation and pressure calibration
x ray diffraction
Angular-dispersive powder X-ray diffraction experiments above pressures of 250 GPa were performed at the Extreme Conditions Beamline (P02.2) at PETRA III
Methods, X-ray diffraction measurements
raman spectroscopy
We have conducted high-resolution Raman spectroscopy measurements of ice X up to pressures of 253 GPa (Fig. 2).
Compression of ice X section, page 1851
density functional theory
Phonon frequencies and IR/Raman intensities were obtained with CASTEP within the density functional perturbation theory framework
Methods, DFT calculations
ice polymorphs
to our knowledge, this is the most dense H2O polymorph ever observed (Fig. 1)
Summary paragraph, page 1853