Phase transition kinetics of superionic H2O ice phases revealed by Megahertz X-ray free-electron laser-heating experiments

Abstract

H₂O transforms to two forms of superionic (SI) ice at high pressures and temperatures, which contain highly mobile protons within a solid oxygen sublattice. Yet the stability field of both phases remains debated. Here, we present the results of an ultrafast X-ray heating study utilizing MHz pulse trains produced by the European X-ray Free Electron Laser to create high temperature states of H₂O, which were probed using X-ray diffraction during dynamic cooling. We confirm an isostructural transition during heating in the 26-69 GPa range, consistent with the formation of SI-bcc. In contrast to prior work, SI-fcc was observed exclusively above ~50 GPa, despite evidence of melting at lower pressures. The absence of SI-fcc in lower pressure runs is attributed to short heating timescales and the pressure-temperature path induced by the pump-probe heating scheme in which H₂O was heated above its melting temperature before the observation of quenched crystalline states, based on the earlier theoretical prediction that SI-bcc nucleates more readily from the fluid than SI-fcc. Our results may have implications for the stability of SI phases in ice-rich planets, for example during dynamic freezing, where the preferential crystallization of SI-bcc may result in distinct physical properties across mantle ice layers.

Fig. 1. Sample configuration and diagnostics used for X-ray heating of H₂O at ~67 GPa.

a Photomicrograph showing the three doughnut-type couplers (Au, Cu, and Ag) in DAC 8, which were used to indirectly heat the H₂O sample during XFEL irradiation. The couplers were imbedded in the Re gasket and insulated from the diamond anvils by a thin layer of H₂O, as illustrated in (b) and (c). Data were collected with the XFEL beam (<8 µm FWHM) aligned to the center of the coupler hole so that the coupler was heated by the tails of the beam. d–f Data from run 533, which was collected from the Cu coupler using 70% X-ray transmission. d Pulse energy as a function of time for the 300 pulses in the train. e SOP spectrogram after fluorescence removal showing the thermal emission from the coupler during the run, where time 0 corresponds to the arrival of the first XFEL pulse. f Temporal evolution of the total SOP intensity and temperature determined from a Planck fit to the thermal emission in a 9.07 µs time window, where the horizontal error bars indicate the bin width and the vertical error bars correspond to one half of the standard deviation confidence from the Planck fit. The temporal resolution of the SOP is not sufficient to resolve temperature oscillations during the heating/cooling process (Fig. 2a); instead, it is sensitive to the hottest part of the run where thermal emittance is the brightest. Source data for panels (d)–(f) are provided as a Source Data file.

Fig. 2. Results of a FEA model used to simulate X-ray heating.

The model simulates X-ray irradiation of the Cu coupler in DAC 8 (run 533). a Temperature evolution at the coupler edge, in the crucible center, and 0.5 µm from the edge of the coupler. Although the coupler temperature undergoes continuous oscillations, these are strongly damped in the center of the crucible and the temperature stabilizes after ~20 pulses. However, temperature oscillations in the H₂O are still significant in the vicinity of the Cu coupler (~500 K at 0.5 µm from the coupler edge). b Temperature distribution in the sample at the time of the 82nd XFEL pulse, showing conditions detected with XRD. Black lines show possible boundaries between ice VII and SI-ice, taken to be at 1500 K or 2000 K. The XFEL beam is incident from below. Temperatures in (a) are reported at the blue, green, and orange points. Source data for panel (a) are provided as a Source Data file.

Fig. 3. X-ray diffraction collected during XFEL irradiation of H₂O at 69.3 GPa.

Data were collected using an Ag doughnut coupler (DAC 8) and an X-ray beam diameter of <8 µm FWHM. a, b Integrated XRD patterns and (c–h) unwrapped (2θ-φ) XRD images collected during irradiation with 300 XFEL pulses at 29% transmission (run 505). In (a), darker shades correspond to higher intensity. In (c–h), lighter shades correspond to higher intensity. Heating results in a splitting of the ice (110), Ag (111), and Ag (200) reflections, which is indicative of two distinct temperatures in the probed volume, predominantly originating from the lack of thermal insulation between the H₂O and the diamond anvils. Panel (a) illustrates how heating occurs during the first ~20 pulses, after which the diffracted signal from H₂O remains essentially constant. In (d), the diameters of the orange circles are unrelated to individual spot sizes. Source data are provided as a Source Data file.

Fig. 4. Integrated patterns from summed images and histograms showing the results of the spot finding approach.

Data are shown for (a) run 505, and for (b) all runs collected from the Ag coupler in DAC 8 where SI was observed. The construction of the histogram for run 505 is illustrated in Supplementary Movie 1. When the Re contribution (estimated from the first pulse pattern) is subtracted from integrated patterns, they are in good agreement with the histogram. Source data are provided as a Source Data file.

Fig. 5. Histograms showing the number of H₂O diffraction spots in runs collected from different samples.

For each sample, histograms were compiled from XRD data collected from pulses 51-300 in runs where SI ice was observed. The width of each bin is 0.01 degrees. Histograms are compared with integrated XRD patterns produced from the summed Adaptive Gain Integrated Pixel Detector images of the hot (pulses 51–300) and cold (pulse 1) sample. The y-axis refers to the histogram plot, and the integrated patterns from the hot and cold sample were normalized to a single arbitrary scaling factor for the figure. The coupler material is specified in each case. The high-angle signal in the DAC 3 and 5 runs are from the hot coupler. With the exception of DAC 1 and DAC 2, in which the coupler was in the form of a dispersed nanopowder, all couplers were of a doughnut-type design. Source data are provided as a Source Data file.

Fig. 6. XRD data collected during X-ray heating of H₂O at 67.3 GPa.

Data were collected during irradiation of the embedded Cu doughnut coupler in DAC 8 using an X-ray beam size of <8 µm FWHM. Integrated XRD data are shown from runs (a) 532 and (b) 536, and single pulse, unwrapped (2θ-φ) XRD images from runs (c) 532 and (d) 536 are shown for a reduced 2θ range. In (a–b), darker shades correspond to higher intensity, whereas in (c–d), lighter shades correspond to higher intensity. The data shown in (a) and (c) were collected using 60% transmission, and the (110) SI-bcc reflection is observed from pulse ~5 onwards (see insert), The SI-bcc diffracted signal is predominantly from several intense spots, such as the one highlighted in (c). The data in (b) and (d) were collected at 100% transmission, and both SI phases (bcc and fcc) are observed. Source data are provided as a Source Data file.

Fig. 7. Average coupler temperatures in experiments where SI ice was observed.

Data points correspond to the average SOP temperature during the run, and the error bars show the standard deviation. Source data are provided as a Source Data file. Thermal emission was not detectable for all samples, and data points are shown for runs in which thermal emission was observed. Due to the low emissivity of H₂O, SOP measures thermal emission from the coupler surface, which does not necessarily correspond to the temperature in the H₂O sample. Our data points are compared to the phase diagram of H₂O reported in previous studies. The text labels indicate the phase stability regions reported in ref. . For simplicity, we do not make any distinction between ice VII and VII’ in the main text. SI-bcc and SI-fcc data points from previous static^,, and shock compression^, studies are included for comparison, as well as ice VII data points from shock compression work^, and the H₂O melting curves from ref. ^,. Ice VII data points from static compression studies are not included to avoid overcomplicating the figure. The data points from ref. . indicate the SI-bcc/ice VII phase line determined on isothermal compression.

Fig. 8. Unwrapped (2θ-φ) XRD images showing evidence of melting of H₂O.

Data were collected using 90% X-ray transmission from the embedded Ag coupler in DAC 4 (36.7 GPa) using an X-ray beam size of <8 µm FWHM (run 917:17). Lighter shades correspond to higher intensity. Source data are provided as a Source Data file. The significant reduction in the intensity of the ice VII/SI-bcc diffracted signal during the run is attributed to melting of H₂O. No evidence of the SI fcc (111) reflection was observed, which would be expected to be present at approximately 18.4°. Unwrapped images from the full run are shown in Supplementary Movie 2.

Fig. 9. Density as a function of pressure for different phases of H₂O.

The data points from this work show the SI-bcc and SI-fcc densities calculated from the histograms in Fig. 5. Source data are provided as a Source Data file. The data point from DAC 9 (34.2 GPa) is included to illustrate the agreement with other data, despite weak evidence of chemical reactivity in this sample. Our results are compared with SI ice data from previous DAC experiments^,,, which are identified as using laser heating (LH-DAC) or resistive heating (RH-DAC) techniques. Ice VII data from ambient temperature^,,, and high temperature DAC experiments, shock compression experiments^,, as well as data from the high temperature fluid, are shown for comparison. In addition to the data points taken from the main paper of Weck et al., which indicate the SI-bcc and SI-fcc densities at the transition temperature, their SI-fcc data point at 57 GPa and 1927 K is also included to indicate the volume of thermally-expanded SI-fcc at higher temperature.