Microscopic dynamics of gas molecules confined in porous channel-like ice structure

For more than a century, the ice polymorphism has been representing a very hot topic for different research fields because of the fundamental role that water takes on in the biosphere as well as in different cosmological contexts.^1–3 Thanks to the continuous development of the investigation techniques, an increasingly larger portion of the water phase diagram becomes experimentally accessible, allowing us to discriminate, to date, three pure forms of amorphous ice,⁴ 21 crystalline ice structures,^5,6 even with the symmetrization of hydrogen bonds⁷ or in the superionic states of H₂O,⁸ and many others that have been computationally predicted.^9–11 In this complex picture, one of these ice forms, ice XVII, that is the emptied counterpart of the high-pressure H₂-hydrate in the C₀ phase,¹² resulted to be a very interesting material not only for its metastability, at ambient pressure and temperatures below 130 K, but also for its porosity, i.e., its ability to absorb and release gas in a surprisingly versatile manner. Indeed, by means of modest gas pressures (up to few bars), ice XVII is able to reversibly absorb not only H₂ but also other atomic and molecular species,^13,14 with relatively fast kinetics and for a theoretically infinite number of absorption/desorption cycles.¹² To date, other ice polymorphs have also been obtained by emptying different high-pressure gas hydrates, but they do not show such a strong porous character.^15–17 Thus, in the ice polymorphism landscape, these unique features of ice XVII, together with the simplicity and environmental friendliness of the chemicals involved (i.e., just simple pure water ice), turned the spotlight on its possible role in the so-called “hydrogen cycle” as a performing nanoporous material for hydrogen storage.¹⁸ Indeed, once the hydrogen is produced, it becomes crucial to have a storage system suitable for a specific application, e.g., small mobile tanks ideal for the automotive applications or large volume reservoirs for stationary storage, even in insular or off-shore environments.^19,20 With the aim of enhancing the stored volumetric energy density, different approaches were deeply investigated in the past, i.e., gas compression, liquefaction, or solid-state, either chemical or physical, adsorption. Among these hydrogen storage systems, hydrogen hydrates, such as clathrates¹ and “filled ice,”^12,21 are both particularly interesting for their storage capacity (e.g., up to 47 g/l for ice XVII), surpassing the record that was determined by the simulated examination of approximately half a million different Metal–Organic Frameworks (MOFs).²² Very recently, a further hydrogen filled ice phase, named C₃,²³ with a very high H₂ content (up to 270 g/l) has been discovered to be stable under extreme pressure conditions (around 400 kbar), demonstrating even more the wide potentiality of this binary system for storage purpose.

In this perspective, ice XVII appears as the most versatile between the hydrate compounds because it can be used as an ice matrix in which it is possible to intercalate different molecules close to ambient pressure, opening in this way the route to various technological applications, among which gas sequestration, purification, and storage are surely the most intriguing. In this work, we present an experimental study on the vibrational dynamics of diatomic molecular gases H₂, N₂, and O₂ absorbed in the nanoporous structure of ice XVII. The investigations, mainly conducted by Raman and neutron spectroscopy techniques, have shed light not only on the quantum dynamics of these systems but also on the peculiarity of this porous medium with respect to other forms of emptied hydrates discovered to date.

In order to produce the porous ice XVII, we started with the high pressure synthesis of the C₀ hydrate, performed at the High Pressure Laboratory of CNR-IFAC (Sesto Fiorentino, Italy). This hydrate is obtained by pressurizing finely grinded D₂O ice powder with H₂ gas at about 4.2 kbar and a temperature of 255 K in a large volume copper-beryllium autoclave. After few hours under this high pressure condition, we were able to recover at ambient pressure the hydrogen C₀ hydrate powder by quenching the autoclave in a liquid nitrogen bath at 77 K and then releasing the gas pressure. The as synthesized sample, named “pristine” in the following, was transferred into cryogenic vials stored in a liquid-nitrogen long-term Dewar. Thanks to the large amount of hydrate powder (about 20 g for each synthesis batch) that we are able to produce, we studied this recovered sample in different ex situ experiments using Raman and inelastic neutron scattering (INS). Given the metastable nature of the recovered sample,²⁴ the whole transfer processes of the sample powder from the Dewar to the measurement cells, for both Raman and INS experiments, were performed without interrupting the cold cryogenic chain in dry nitrogen atmosphere to prevent water vapor condensation on the cold sample surfaces. The transformation from H₂ hydrate in the C₀ phase into empty ice XVII was obtained by applying a so-called annealing treatment, i.e., keeping the sample at 110 K under dynamic vacuum for a maximum of 2 h. This was directly performed in the measurement cells (see Fig. 1), which are in thermal contact with an He cryostat and are provided with a gas capillary connected to a low pressure gas circuit, thus allowing not only to evacuate the sample volume but also to dose the different gas species investigated (i.e., H₂, N₂, and O₂) up to a nominal pressure of 3 bar. We chose to study deuterated ice in order to better compare the Raman data with INS, which requires the minimization of the incoherent contribution from the scattering of the water skeleton to highlight guest molecules’ self-dynamics.

The Raman experiment was performed by means of a customized spectrometer at the CNR-IFAC Raman Spectroscopy Laboratory (Sesto Fiorentino, Italy). The optical path, mounted in back-scattering configuration, is equipped with a red line of a He–Ne laser (633 nm) as an excitation source. The scattered light, collected using a 50 mm-focal length lens, travels through a complex system of filters composed of a spatial one, which reduces the divergence of the collimated beam, and three spectral SuperNotch filters, which allow to effectively remove the spurious stray-light contribution down to 25 cm⁻¹. After the filter system, the beam is focalized onto the entrance slits of the first stage of a SPEX monochromator, which is coupled with a 1024 × 128 px ANDOR CCD detector. The absolute calibration of the energy scale of each spectrum was performed before each experimental campaign by a measurement of the Ne lamp spectrum in the same spectral window used for the sample investigation (i.e., 25–375 cm⁻¹), obtaining a final nominal resolution of 0.5 cm⁻¹. The INS measurements were performed by means of the neutron inverse-geometry time-of-flight spectrometer TOSCA-II,²⁵ installed at the ISIS Neutron and Muon Source (Rutherford Appleton Laboratory—STFC, Didcot, UK), which is characterized by energy resolution 1.5% ≤ ΔE/E_i ≤ 2.5%, with E_i being the incoming neutron energy. By applying the correction for the kinematic factor and the normalization for the incoming neutron flux, raw time-of-flight data have been transformed into energy-transfer spectra.

The low-energy region of the Raman spectra, i.e., 0–400 cm⁻¹, of the sample here investigated contains different vibrational modes belonging to both D₂O ice structure and gas species possibly absorbed. Thanks to the implemented Raman setup, we were able to collect a wide portion of this energy region (25–375 cm⁻¹) all at once, measuring, in the same spectral window, all the modes of interest, and allowing in this way their direct comparison as the filling is performed. The list of the measured samples, together with the relative thermodynamic condition, is reported in Table I. The low-energy Raman spectra of the empty ice XVII sample superimposed on the spectra of the H₂, N₂, and O₂ refilled samples are reported in Fig. 2. In this spectral window, we can always recognize a quite large Raman band ranging from 175 to 325 cm⁻¹ originated by the collective motion of the D₂O molecules of the host structure, i.e., the lattice phonon bands.¹² When dosing the different gas species, new bands, originated from the roto-vibrational motions of the guest molecules confined in the ice structure, start to appear in distinct spectral regions. In particular, the strong band at 325–375 cm⁻¹ (see left panels of Fig. 2) is due to the light H₂ molecule, while the large band at lower energy values, i.e., 35–125 cm⁻¹ (see the central and right panels of Fig. 2, respectively), is due to the presence of the heavier N₂ or O₂ molecules. Together with the appearance of these new Raman bands, we can note, especially at the higher temperatures for the heavier guest species (i.e., N₂ and O₂), a slight distortion of the lattice phonon band of ice XVII after the absorption of the gas molecules, with the growing of the left shoulder of the main spectral feature centered at about 220 cm⁻¹ and a less evident change also in the right side of the same band. This fact confirms that guest molecules were effectively intercalated in the ice porous matrix, causing a measurable change in its collective dynamics due to the distortion of the crystalline structure, with a higher degree as the increase of the dimension of the guest and the strength of the van der Waals guest-host interactions.

In order to better assign the vibrational modes of the confined guest molecule, we need to maximize the degree of filling of ice XVII by the three different gas species. With this aim, we also performed Raman measurements under moderate pressure conditions (between 0.4 and 3.0 bar, depending to the dosed gas), in order to have a supersaturated atmosphere in the optical cell, but at a higher value of temperature, i.e., in the range 50–85 K, avoiding the formation of their condensed phase, in particular for N₂ and O₂ gases. As it appears evident from bottom panels of Fig. 2, in all the three cases, the very narrow spectral lines superimposed to the quite large bands demonstrate that the molecular species are present in the form of both free gas phase and confined in the ice XVII structure, respectively. Indeed, the confinement and, as a consequence, the anisotropic potential on the guest molecules imposed by the spiraling water molecule arrangement^13,28 generate a perturbation of the molecular quantum levels here involved in the Raman-active transitions, giving rise to very large and structured bands, centered around the free gas lines. This is the case, for example, of the hydrogen filling (the left bottom panel of Fig. 2), where the large band centered at 350 cm⁻¹ is due to the single-H₂ rotational transition S₀(0), in which the guest–host interaction partially removes the degeneracy of the J = 2 levels with respect to the quantum number m. A similar behavior is recorded also for the nitrogen and oxygen filling (the central and right bottom panels of Fig. 2, respectively). Here, the rotational lines are located at lower frequencies and have a strongly reduced separation as a function of quantum rotational number J with respect to the hydrogen case. This is clearly explainable by considering, in the first approximation of a rigid diatomic molecule, the equilibrium rotational constant $Be=h8π2μre2$⁠, with h being the Plank constant, r_e being the equilibrium distance of the two nuclei, and μ being its reduced mass.²⁹ Thus, due to the higher mass of the nitrogen and oxygen molecules, their B_e values were reduced by a factor of 14 and 16 with respect to the hydrogen molecules, respectively, allowing to collect several rotational lines in the range 25–130 cm⁻¹. On the contrary, this reduced separation causes a strong mixing of the rotational components when the confinement is considered. In order to collect a clean spectrum of the sole confined molecules in the ice structure, we have performed another set of measurements partially removing the gas inside the cell up to the disappearance of the free gas lines. The resulting spectra are reported in Fig. 3, with some spectral details listed in Table II. Concerning the Raman D₂O lattice bands, we can just recognize minor changes of their shapes for the two confined guests considered. This can be rationalized, in the reasonable hypothesis that both nitrogen and oxygen molecules occupy the same crystallographic sites,¹³ by considering the similar nature of the guest–host interaction potential for the two guest molecular species, thus causing a similar distortion of the water molecule crystalline structure. From the comparison of the low-frequency band in the oxygen- and nitrogen-filled ice XVII, it is quite evident that the higher intensity in the former case (see also Table II) denotes a higher capacity of storing oxygen with respect to nitrogen at thermodynamic equilibrium of 85 K. Moreover, we can clearly see a similar spectral shape of the bands for the two kinds of confined guest, with the main spectral features of the heavier species, i.e., oxygen, red-shifted by a factor 1.2–1.3, a value similar to the oxygen-to-nitrogen molecular mass ratio, which is 1.14. Even if this fact confirms the rotational nature of the mode that originates this Raman band, the width of the latter suggests that the rotation of the molecules is strongly hindered by water confinement, in a similar way to what happens in the case of solid nitrogen³⁰ and solid oxygen.^27,32 In particular, the latter shows two distinct bands placed in the same spectral region (see Table II), with the lower-frequency one showing higher intensity and lower width, as in our case. In this sense, the presence of two main peaks can be a spectral sign of the two degrees of freedom with which we can describe the librational motion of the confined guest molecules, that is the librations of O₂ and N₂ molecules around the two axes perpendicular to their molecular axis. Being these bands even larger with respect to those measured for the solid diatomic phases,²⁷ we can rationalize them with a stronger perturbation of the librational motion induced by water confinement and a possible orientational disorder of the molecule, which can also give rise to the weak wing at low frequencies present in our data for both the confined species.

Given the peculiar nature of the neutron probe interacting with condensed matter, the INS technique allows us to effectively measure the dynamics of the guest molecular species confined in the ice XVII porous structure. Moreover, the absence of the selection rules also allows us to measure Raman-inactive transitions involving, for example, rotational, with ΔJ = 0, 1, and center-of-mass vibrational (i.e., rattling) motion of the guest molecule. We chose to use a deuterated water skeleton to highlight the vibrational dynamics of the guest, especially in the case of nitrogen and oxygen, reducing in this way to a minimum the contribution of the ice signal. Once the pristine sample powder was cryo-loaded into the TOSCA container, we applied a preliminary annealing treatment to obtain empty ice XVII. After that, for each studied species, we dosed the different types of guest at temperatures and pressures so as to avoid the gas condensation in the sample container. Subsequently, we cooled again the sample down to low temperature, i.e., 15 K for the hydrogen-filled samples and 20 K for the nitrogen- and oxygen-filled samples, for the actual measurements. Between two different gas fillings, the same annealing procedure was performed to completely evacuate the guest molecules from the ice structure. The details of the gas dosing during the TOSCA experiment are reported in Table III. The full spectra, as collected by means of the TOSCA spectrometer, are reported in Fig. 4. As it appears evident, in the case of hydrogen loading, the INS signal is dominated by the strong incoherent contribution of the H₂ guest molecule, with the overall intensity that increases as the content of confined hydrogen grows. In these spectra, we can simply recognize the J = 0 → 1 rotational triplet of para-H₂ at 14 meV and the J = 1 → 2 and J = 1 → 3 rotational bands of ortho-H₂ at 29 and 73 meV, respectively. The other spectral features were successfully assigned to the combination of the H₂ center-of-mass vibrational motion (i.e., rattling) with the elastic and inelastic rotational transitions of both spin isomeric species.³⁵ On the contrary, in the case of nitrogen and oxygen loading, due to their weak scattering power, the INS signal is dominated by the scattering of the ice matrix, with the guests that essentially cause only minor spectral changes in the low-energy part of the spectra, i.e., below 20 meV. However, in order to better highlight the signal from the confined guest molecules, we needed to subtract the blank sample, i.e., the empty ice XVII spectrum, which could not be neglected in all the three cases. The H-projected density of phonon state of pure ice XVII was already studied in a previous work,²⁴ and so, here, we can simply point out the main spectral features: acoustic and optic phonon bands (0–40 meV) and librational bands (45–95 meV), but with small spectral differences due to both the higher coherent scattering contribution of our deuterated sample and the mass differences which placed the main features at a lower exchanged energy with respect to the mentioned H₂O case. Indeed, the phonon and librational bands in the deuterated case are red-shifted by a factor of 1.1 and 1.4, respectively, which are in good agreement with the expected frequency ratio calculated by considering, in the simple harmonic approximation, the square root of the molecular mass ratio (1.05) and the atomic mass ratio (1.41) for the phonon and librational case, respectively. This fact is similar to what has been recorded for both the amorphous³⁶ and the crystalline^24,37 ice phases.

Concerning the microscopic dynamics of the hydrogen molecule inside the D₂O ice XVII channel structure, once the reference spectrum (i.e., empty D₂O ice XVII and sample container) is subtracted, neglecting the coherent scattering contribution, the resulting spectrum can be approximated with the convolution product between the self-dynamical structure factor, modeling the H₂ center-of-mass vibrations, and a comb of Dirac delta functions, weighted by a suitable molecular form factors,^38,39 modeling the H₂ rotational transitions, as efficiently performed in Ref. 35. Taking this fact into account, also in the current spectra, we can point out the existence of at least two kinds of guest vibrational modes inside the channel structure: the center-of-mass vibrational (rattling) mode together with the rotational modes. In the present work, we have performed a further study involving three different levels of filling with hydrogen gas and recording spectra at low temperatures after each dosing. On the basis of the assignment performed in Ref. 35, we have extracted the spectra of the center-of-mass excitations by subtracting from the experimental spectra the pure rotational excitations J = 0 → 1, previously identified by means of fitting procedure involving Voigt-shaped peak functions. The resulting curves, reported in Fig. 5, are mainly due to the rattling motion of ortho-H₂, even if the very weak shoulder, centered at about 17 meV, can be assigned to the combination of the rattling and the J = 0 → 1 rotational mode of the para-H₂ species (see the mode denoted by a″ in Ref. 35). We can clearly recognize two well-distinct bands centered at about 7 and 20 meV, already successfully assigned to the vibration of the H₂ center-of-mass in the directions along and across the axis channel of the ice XVII crystallites, respectively.³⁵ These bands appear quite large, and so we can hypothesize the presence of other sub-components coming from the removal of the m degeneracy by the anisotropy of the guest–host interaction potential. As for the shape of the bands, only the low-frequency one is affected by the degree of filling. Recalling the spectrum of the center-of-mass velocity autocorrelation function, computed for half and full filling (see Fig. 3 of Ref. 35), we see a modest agreement with the observed behavior, i.e., both spectral features of the larger filling get blue-shifted. However, this is true only for the low-frequency center-of-mass translational band and can be rationalized, if an increasing filling is considered, with the decrease in the H₂-molecule freedom to rattle in the direction along the channel axis, due to the increasing presence of other H₂ neighboring molecules, causing in this way a frequency hardening of the vibrational mode. On the contrary, the H₂ rattling in the directions lying in the plane perpendicular to the channel axis is insensitive to the presence of neighboring H₂ molecules, so no change in the spectral shape of the high-frequency center-of-mass translational band is recorded as the filling increases. Concerning the intensity of such bands, it generally grows with the filling as expected, even though some inconsistencies are recorded for the band centered at about 7 meV and for the 20 meV band that appears more insensitive to the hydrogen filling than the former. In principle, also looking at the intensity of the rotational bands J = 0 → 1 and J = 1 → 2 centered at 14 and 30 meV, respectively, we cannot exclude a possible effect on the intensity of the rattling bands caused by ortho-to-para H₂ conversion. However, given the slow kinetics of this phenomenon in a similar hydrate system (e.g., the sII H₂ clathrate hydrate⁴⁰) with respect to the duration of the present measurements, this phenomenon can explain these inconsistences only in part. In our case, it is likely that the background modeling at low energies is not perfectly accurate, and we would need to improve the description of the diffusion dynamics of the guest molecules inside the ice XVII channel-like structure with a line-shape corresponding to a more realistic diffusional model. This would also help to better interpret the different intensity behavior of the two bands as a function of the guest filling. However, in order to investigate this kind of diffusional dynamics, it would be necessary to measure the quasi-elastic neutron spectrum of this system, although this task is beyond the scope of the present work.

For the filling with the other two types of gases, i.e., nitrogen and oxygen, the resulting density of states in the lattice mode region (i.e., 0–40 meV), as extracted from the TOSCA INS data following a similar procedure extensively reported before (see the Supporting Information section of Ref. 24), is plotted in Fig. 6 for the two detector banks, i.e., forward (F) and backward (B) scattering. Here, given the non-negligible coherent contribution to the scattering with respect to the incoherent one from both the deuterated ice skeleton and the guest species, we cannot simply sum the signal collected by the two different detector banks because they cover different kinematic trajectories in the (Q, ω) plane.²⁵ In general, for the two fillings, in both scattering geometries, we can interpret the spectral changes from the empty to the re-filled ice structure as the ice lattice-mode modifications induced by the presence of the guest molecules. In particular, the differences are really tiny in the optical phonon region (i.e., 19–40 meV), with a slight downshift of such a band and no change of the spectral shape caused by the presence of guest molecules. This fact corresponds with no major change in the Raman translational band from empty to filled ice XVII (see Fig. 2). These minor changes are also verified for the D₂O librational bands, as it appears evident from central and bottom panels of Fig. 4, where the presence of the guests only causes a weak intensity increase in the 60–75 meV exchanged energy region. On the contrary, in the acoustic region (i.e., 0–19 meV), for both guest species, we detect the appearance of a peak around 2.5 meV, which results more intense in the case of the oxygen filling with respect to the nitrogen one, especially in forward scattering, a shoulder at about 7 meV and a strong peak at 14 meV. The second and third spectral features in the acoustic region, which are weakly dependent on the type of guest, can be assigned to the hindered rotational motion of the N₂ and O₂ molecules inside the channel-like ice structure, in good agreement with what has been found in the present Raman experiment (see Fig. 3). However, the lowest-energy feature at 2.5 meV might be interpreted as a resonant effect due to guest–host coupling, which causes a perturbation of the acoustic dispersion branch with a magnitude that depends on the strength of coupling. A similar effect has been measured for the clathrate hydrates, with the appearance of a similar feature at about 2.5 meV in the density of states of nitrogen sII,⁴¹ or in the case of xenon sI clathrate.⁴² However, in order to better address this assignment, we would need to collect new inelastic/quasi-elastic data by another spectrometer with a better resolution in the energy-transfer region below 3 meV.

Among the various ice polymorphs, there are only two other cases of pure ice phases obtained by evacuation of gas hydrate compounds, other than ice XVII, to be considered as viable candidates for porous ice phases. The first discovered is the ice XVI, which is the emptied counterpart of sII structure Ne clathrate.¹⁵ This empty-cage structure is particularly interesting, even from a technological point of view, because it is able to host hydrogen molecules under high pressure conditions (i.e., around 1–3 kbar) and shows a porous character. Indeed, it has been demonstrated that ice XVI is able to absorb He gas¹⁶ but at pressure around 1 kbar, thus partially hindering its application for gas storage near standard conditions. Recently, the hydrogen filled ice C₂ has also been successfully emptied,¹⁷ demonstrating the effectiveness of this route to obtain pure cubic ice I. However, to our knowledge, no one has demonstrated the capability of these two ice forms to repeatedly absorb and desorb hydrogen and other type of gases near ambient pressure condition.

While for ice XVI, no attempt has been made to date, we were able to test the possible porosity of cubic ice I on a sample obtained following an alternative synthesis route,⁴³ that is by heating recovered ice XVII crystallites above 140 K at room pressure. After this thermal treatment, performed in the Raman gas cell (see Sec. II), the sample, fully transformed in ice I_c, was cooled to 85 K and then pressurized with both hydrogen and nitrogen gas up to a pressure of 3.0 and 2.0 bar, respectively. The resulting Raman spectra are shown in Fig. 7. If an absorption process was recorded, we would expect some signatures on the spectral features of both guest and host molecules, i.e., a large and structured rotational or librational Raman band of guest molecule and some spectral changes in the lattice phonon band of the ice matrix (see Sec. III A). In the case of the hydrogen dosing, we can clearly see the presence of the strong Raman band S(0) at about 354 cm⁻¹ originated by the quantum rotational transition (ν = 0, J = 0) → (ν = 0, J = 2) of the hydrogen molecules inside the optical cell. The width of that band clearly indicates that molecules rotate freely, so only the free gas phase is detected. Moreover, the shape of the lattice phonon band between 200 and 300 cm⁻¹ of the pressurized sample is very similar to the one under dynamic vacuum, demonstrating no distortion of the ice crystalline lattice due the presence of hydrogen gas trapped into the ice structure. Also in the case of the nitrogen-dosed experiment, the presence of free gas in the cell is shown by several narrow rotational lines in the low energy region of the spectrum (i.e., 25–125 cm⁻¹). Moreover, no modification of the lattice translational band of cubic ice is detected between the nitrogen-pressurized and under vacuum ice I_c, thus demonstrating again that cubic ice does not manifest a porous character, at least in the low pressure regime.

The work here presented showed the ability of ice XVII to absorb and release different types of gas in a very versatile manner, allowing us to effectively investigate the confined dynamics of the guest molecules in a molecular solid matrix. The three types of guest species investigated here present peculiar spectroscopic features in the low-energy region, scarcely studied near room pressure to date, which have been effectively assigned to the rotational and rattling motion of the H₂ molecules in a confined environment, while, for the heavier N₂ and O₂ molecules, the increased interaction with the ice skeleton strongly hinders their rotational motions, which turn out to be, in fact, librational ones. The microscopic dynamics of the confined guest molecules investigated here also suggests that both nitrogen and oxygen molecules occupy the same crystallographic sites in the ice unit cell, but this needs to be directly confirmed by new diffraction measurements on nitrogen-filled ice XVII, being only oxygen-filled ice XVII structurally resolved to date.¹³ Moreover, the higher content of oxygen trapped inside the ice XVII matrix with respect to the nitrogen case at thermodynamic equilibrium (85 K) suggests slower diffusional dynamics of the heavier species along the spiraling channel-like structure of ice XVII, but, also in this case, new investigations are needed. This fact could potentially explain why one needs quite a high annealing temperature (up to 120 K) to completely evacuate the initial C₀ hydrate sample as-synthesized,¹⁴ which can also absorb spurious molecular oxygen during the recovery at 77 K.

In a more general ice polymorphism perspective, by comparing with other metastable ice phases originated by emptying hydrates, i.e., ice XVI and ice I_c, ice XVII demonstrates to be unique in terms of porous character, at least in the low pressure regime, which is the most relevant for technological applications. Probably, the difference of these two forms of ice with respect to ice XVII resides in the very open structure of the latter, made of quite large open channels (about 6 Å diameter), which can host guest molecules in a spiraling geometry along the axis channel and are characterized by a larger diffusion coefficient of guest molecules even though they exhibit a reduced dimensionality. As a matter of fact, the out-diffusion of Ne molecules from the sII clathrate structure required quite a higher temperature and a longer time of pumping (about 5 days at 142 K¹⁵) with respect to ice XVII (few hours at 110 K). This is probably due to the pentagons of the clathrate small cages that strongly hinder the guest diffusion mechanism. On the other hand, the lack of porosity at modest pressures of cubic ice I_c is explainable looking at the low-pressure portion of the H₂–H₂O phase diagram (i.e., below 1 kbar), where the stable phase is the normal hexagonal ice I_h in simple coexistence with hydrogen gas.⁴⁴ This compound presents a low-hydrogen concentration, comparable with that pertaining to the liquid phase of the H₂–H₂O mixture under similar pressure conditions, thus excluding the porous behavior near standard ones. So, ice I_h being characterized by a water molecule arrangement very similar to ice I_c, the expected non-porous behavior of the latter is in agreement with what we observed in the present work (see Sec. III C). In this scenario, there is a further gas hydrate phase to be considered, that is the filled ice C₁. This hydrate is stable at high pressures in the presence of hydrogen (above 7.5 kbar)^21,45 and is characterized by a host lattice with a linear channel-like structure that closely resembles that of ice II. This direct link between the phase diagrams of the binary compound H₂–H₂O with that of pure ice suggests the filled ice C₁ as a viable candidate for new emptied hydrates with possible porous capability. However, the recovery and handling of samples synthesized at such a high pressure represents a technological limit that has yet to be overcome.

Finally, these fundamental studies about emptied hydrates, such as ice XVII, as innovative porous media could open the way to many possible applications, especially in the so-called “green transition.” The readiest, at least to be proposed at a low “technology readiness level” (TRL), is in the hydrogen storage field, given the high H₂/H₂O molar fraction (up to 0.5) that ice XVII is able to reach. On the other hand, the capability to absorb different kinds of light gases, as well as the probably non-porous behavior with heavier gases, make us think of a possible use of ice XVII as a molecular sieving to be applied to gas sequestration and/or purification, for example in the gray-hydrogen production.¹⁹ In this perspective, further studies will need to determine to what cases the porous behavior can be extended.

We acknowledge ISIS for the provision of neutron beam time and thank ISIS sample environment staff for their fundamental technical support. We thank L. Ulivi (CNR-IFAC) for the fruitful discussions.

This research was funded by the European Union – NextGeneration EU, within PRIN 2022, PNRR M4C2, Project “E-ICES” 2022NRBLPT_PE3_PRIN2022 (CUP: B53D23004390006). We also acknowledge the financial support of the DIITET-CNR with the projects “TIRS” (DIT.AD022.180, CUP: B55F20002150001) and “STRIVE” (DIT.AD022.207, CUP: B53C22010110001) and the financial support of Fondazione CR Firenze with the project “Grandi Attrezzature” (Grant No. 2019.0244). This research was also funded by the European Union – Next Generation EU from the Italian Ministry of Environment and Energy Security POR H2 AdP MMES/ENEA with involvement of CNR and RSE, PNRR - Mission 2, Component 2, Investment 3.5 “Ricerca e sviluppo sull’idrogeno” (PRR.AP015.017 H2 - AdC ENEA/CNR POR IDROGENO, CUP: B93C22000630006).

The authors have no conflicts to disclose.

L. del Rosso: Conceptualization (equal); Data curation (equal); Investigation (equal); Project administration (equal); Resources (equal); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). D. Colognesi: Conceptualization (equal); Data curation (lead); Investigation (equal); Project administration (equal); Resources (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). A. Donati: Methodology (equal); Resources (equal); Writing – review & editing (equal). S. Rudić: Data curation (equal); Investigation (equal); Methodology (equal); Resources (equal); Writing – review & editing (equal). M. Celli: Conceptualization (equal); Data curation (lead); Investigation (equal); Project administration (equal); Resources (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request. Neutron raw data are available from the ISIS depository^33,34 after an embargo time.