Raman studies of hydrogen trapped in As₄O₆·2H₂ at high pressure and low temperature

Hydrogen is an emerging fuel of the future. It is ecological, as only water vapor is formed after its combustion, and very light offering high density of energy if a proper method of storage is applied.¹ Nonetheless, there are still many challenges related to cheap hydrogen production and efficient storage. The basic research into intermolecular interactions of H₂ molecules with host materials to elucidate factors responsible for their strength and directionality and, in turn, to help scientists and engineers design suitable hydrogen storage materials is necessary. Hydrogen inclusion compounds are good model compounds for such research. A few years ago, we discovered an arsenolite–hydrogen inclusion compound As₄O₆·2H₂ that forms above 1.5 GPa and characterized its structure and properties using x-ray diffraction, Raman spectroscopy, and density functional theory (DFT) computations.² The compound crystallizes in the cubic $Fd3¯m$ space group (no. 227, origin choice 2) with As₄O₆ and H₂ molecules each centered on special Wyckoff positions. Arsenolite molecules sit at the 8b site with $4¯3m$ (T_d) point group symmetry, while hydrogen molecules are located at the 16c site with $3¯m$ (D_3d) point group symmetry (see Fig. 1). We observed that the H₂ Q₁(1) vibron is softened in the inclusion compound compared to fluid H₂ over the entire pressure range studied, and the vibron continues to soften with increasing pressure. Interestingly, the vibron was observed to split into three components up to 4 GPa and into two components up to 10 GPa even though structural and computational results suggest that H₂ molecules occupy only one Wyckoff site.

Multiple and/or split hydrogen vibrons have been observed for related inclusion compounds and may indicate a variety of physical phenomena. For instance, clathrate hydrates of hydrogen exhibit multiple H₂ vibrons due to the fact that various numbers of H₂ molecules occupy large and small cavities therein and, additionally, each of these bands are split to reflect the population hydrogen’s two spin isomers.^3,4 Owing to the fact that ¹H nuclei are fermions, the total wavefunction of a hydrogen molecule must be antisymmetric with respect to nuclei permutation. This, combined with the symmetry of the electronic part of the wavefunction, leads to the fact that hydrogen at room temperature (RT) is composed of a mixture of two spin isomers that cannot interconvert quickly into one another. ortho-¹H₂ is the isomer with the symmetric nuclear spin wavefunction (I = 1) and occurs only in rotational states described by odd rotational (J) quantum numbers, whereas para-¹H₂ exhibits an antisymmetric nuclear spin wavefunction (I = 0) with even J. The situation is different for deuterium, in which the ²H nuclei are bosons. The ortho-²H₂ isomer exhibits a symmetric nuclear spin wavefunction (I = 0 or 2) and even J, whereas para-²H₂ possesses an antisymmetric nuclear wavefunction (I = 1) and odd J. The equilibrium distribution of ortho to para H₂ is a result of an interplay between the degeneracy of nuclear spin and rotational wavefunctions as well as Boltzmann factor. In summary, the ortho–para ratio for ²H₂ (deuterium) at RT is 2:1 and increases upon cooling, whereas the ratio at RT for ¹H₂ (protium) is 3:1 and decreases upon cooling.^3,5 It is noteworthy that the ortho–para conversion is forbidden by selection rules and requires the exchange of angular momentum, which is typically a very slow process. The presence of an exchange catalyst, containing paramagnetic oxides like commercially available Fe₂O₃·2%H₂O or CrO·SiO₂, is often used to increase conversion rates.⁶ 

Other examples of systems with multiple or split H₂ vibrons include compounds such as Ar(H₂)₂,⁷ Xe(H₂)_x,^8,9 SiH₄(H₂)₂,^10,11 GeH₄(H₂)₂,^12,13 (H₂S)₂H₂,¹⁴ (H₂O)₂H₂,¹⁵ Kr(H₂)₄,¹⁶ and H₂ dissolved in solid Ne.¹⁷ The complex Raman spectra observed in compounds such as SiH₄(H₂)₂, which exhibits at least seven vibrons, were explained based on slight perturbations of freely rotating H₂ molecules leading to overall lowering of the crystallographic symmetry,¹¹ similar to the statistical distribution of multiple local environments for H₂ dissolved in rare gas solids.¹⁷ In other compounds, such as high-pressure (HP) (H₂S)₂(H₂), multiple hydrogen vibrons reflect different crystallographic sites for hydrogen molecules.¹⁴ For the case of Xe(H₂)_x, a hydrogen 3 × 3 × 3 supercell was postulated.^8,9,14

Herein, we present a detailed high-pressure (HP) Raman study of the arsenolite inclusion compound spanning from room temperature (RT) to low temperature (LT) both with hydrogen and deuterium to elucidate the physical origins of the observed hydrogen vibrons in As₄O₆·2H₂. This paper is organized as follows: Sec. II explains the experimental and computational approach that we have applied, Sec. III describes the results from our studies, and, finally, Sec. IV contains discussion and conclusions.

Symmetric diamond anvil cells (DACs) typically equipped with diamonds having 400-µm culets together with Re gaskets pre-indented to a thickness of 50 µm–60 µm and with laser-drilled 250-µm circular holes were used. The DACs were capable of in situ high-pressure and low-temperature Raman measurements. Thin pellets of arsenolite powder were placed in the gasket hole alongside with 2–3 ruby chips used for pressure determination from ruby fluorescence shift.^18,19 The pressure scale of Dewaele et al. along with the temperature corrections from Datchi et al. was applied, and fitting was carried out using the T-Rax program.^20,21 The DACs were subsequently gas-loaded with protium ¹H₂ or deuterium ²H₂, and Raman measurements were carried out. The samples were excited using a 532-nm diode laser. The laser was focused onto the sample using a 20× (10× in the case of LT measurements) objective lens. The backscattered light was detected using a Princeton Instrument spectrograph SP2750. An 1800-grooves/mm grating was used to disperse the Raman light onto a liquid-nitrogen-cooled CCD, which enabled a spectral resolution of ∼1 cm⁻¹. Initial experiments indicated that the As₄O₆·2H₂ inclusion compound is sensitive to intense laser light by the observation of elemental arsenic formed upon irradiation via the reduction of arsenolite by hydrogen. Consequently, subsequent measurements were performed with a laser power near 1 mW to record Raman spectra with a sufficient signal-to-noise ratio and to avoid sample damage. Neon emission lines were used to calibrate the spectrometer. The spectra were fitted with Lorentzian peaks using the fityk 1.3.1 program.²² In some LT runs, Fe₂O₃ powder, left in the air for a couple of days to absorb moisture, was also loaded into DACs to help facilitate the ortho–para conversion of hydrogen. Computations of Raman modes were carried out according to the procedure outlined in Ref. 2.

HP, LT powder x-ray diffraction measurements on the arsenolite hydrogen inclusion compound were carried out at beamline 16-BMD, HPCAT, at the Advanced Photon Source synchrotron. A helium-cooled cryostat was used with DACs equipped with 600-µm culets to help minimize the pressure variation with temperature. Diffraction data were collected with a MAR 345 image plate. The raw images were integrated with the Dioptas 0.5.0 software, and Pawley fitting was carried out using GSAS-II.^23,24

Ab initio molecular dynamics simulations for the As₄O₆·2H₂ inclusion compound were performed with the NPT (N is the number of particles, P is the pressure, and T is the temperature) ensemble implemented in Vienna ab initio Simulation Package (VASP) code with Langevin dynamics.^25,26 The all-electron projector-augmented wave (PAW) potential was adopted with the PAW potentials taken from the VASP library, where 4s²4p³, 2s²2p⁴, and 1s¹ are treated as valence electrons for As, O, and H atoms, respectively.²⁷ The plane wave was expanded to an energy cutoff of 400 eV, and Brillouin zone sampling with a 2 × 2 × 2 Monkhorst–Pack k-mesh was employed.²⁸ Exchange and correlation effects were treated in the Perdew–Burke–Ernzerhof parameterization of the generalized gradient approximation.²⁹ Dispersion interactions were accounted for using Grimme’s D3 correction with standard damping function—D3(zero).^30,31 Trajectories of atomic movements were visualized using the VMD software.³² 

Variable-temperature and pressure Raman spectra of powdered As₄O₆·2H₂ are shown in Figs. 2 and 3. We found that the formation rate of the arsenolite–hydrogen inclusion compound is strongly dependent on the pressure at which the compound is formed. We were able to observe its formation for protium only above 1.5 GPa. The reaction takes tens of minutes to complete at this pressure, while at 1.8 GPa, it is completed within a few minutes (see Fig. S1 for the evolution of spectra as a function of time at ∼1.5 GPa). The formation of the inclusion compound is evidenced by a discontinuous change in the frequencies of As₄O₆ molecular vibrations and by the appearance of H₂ vibrons in the Raman spectrum. The evolution of Raman spectra with pressure at room temperature, depicted in Figs. 2(a) and 3(a), confirms our earlier observations from the Raman spectra recorded for the single crystalline sample of the inclusion compound. However, careful acquisition of spectra from a thicker powder sample and subsequent fitting of the spectra revealed the presence of three ¹H₂ vibrons over the entire pressure range studied [up to 24.3(2) GPa] for the inclusion compound (see Figs. S2 and S3 for the frequency dependence of modes with pressure). Our interpretation of the previous Raman spectra pointed to the presence of three vibrons below ∼5 GPa and of two vibrons above this pressure.² For clarity of presentation and discussion, the vibrons will be hereafter referred to as ν₁, ν₂, and ν₃, and their numbering corresponds to increasing wavenumber. It is noteworthy that (1) all three components soften with pressure, while the opposite trend is observed for fluid/solid H₂; (2) the frequency difference in between ν₁ and ν₂ decreases with pressure, and these bands merge above ∼8 GPa (above this pressure, fits of the spectra with two peaks are significantly better than with one peak only); (3) ν₃ becomes very broad with diminished relative intensity at high pressure, and its frequency dependence with pressure deviates from the other two bands. The last two observations explain why we missed the presence of all three components at higher pressures in our initial study on As₄O₆·2H₂.

In order to further understand the vibrational spectra observed in the inclusion compound, we carried out LT Raman measurements with protium, as shown in Figs. 2(a), 2(b), 3(a), and 3(b). Starting in the lower-frequency region below 1000 cm⁻¹, vibrational modes associated with the arsenolite host lattice are clearly identified by comparison with spectra from the empty structure and with DFT computations. Given that all Raman measurements were performed in excess fluid hydrogen, roton excitations [e.g., S₀(0) and S₀(1)] from the pure fluid (or solid) are also observed in each spectrum, but are easily distinguished by comparing the spectra of pure hydrogen at the same conditions. With decreasing temperature, all Raman peak widths decrease as vibrations become more localized. Below ∼200 K, the rotational fine structure becomes apparent within the J = 0 → 2 and J = 1 → 3 manifolds near 350 cm⁻¹ and 600 cm⁻¹, respectively, and these bands broaden with increasing pressure. By 78 K, the splitting of these bands is significant [denoted by asterisks in Fig. 2(b)] and indicates at least partially hindered rotation of hydrogen molecules within the inclusion compound. These quantum rotor excitations are not captured by static phonon computations. Notably, computations do predict two “librational” modes with symmetries T_2g and E_g, in which hydrogen atomic displacements dominate the overall character. These modes might be related to the two broad bands observed at ∼130 cm⁻¹ and 340 cm⁻¹ at 1.3 GPa and LT based on their appearance upon formation of the inclusion compound and their extreme pressure dependence, which agrees with the calculations.

Turning to the high-frequency region, all vibron peaks become sharper as temperature decreases (see Fig. S4), and the ν₁ vibron clearly exhibits a shoulder on the high-frequency side indicating that it is, in fact, comprised of two very closely spaced ν_1a and ν_1b peaks. Given that DFT computations also predict two closely spaced Raman vibrons, one of which is of A_g and the other of T_2g symmetry, it appears that the symmetry of the crystal structure explains the origin of these bands (see Tables S1 and S2). The relative intensities of the three vibrons also reveal distinct trends with temperature. At a constant pressure of 3.6(3) GPa, the relative intensity ratio of ν₁ to all other vibrons (defined as the area of ν_1a and ν_1b over the total area of all vibrons) varies significantly, while the relative intensity ratio of the ν₂ vibron to all other vibrons remains constant at ∼0.25 upon cooling from RT to 80 K. The relative intensity of the ν₃ vibron diminishes from 0.36 at RT to 0.04 at ∼80 K (see Figs. 4 and S5 for an extended version including relative fluid/solid H₂ roton intensities). The relative intensities of the ν₁:ν₃ vibrons recorded at RT do not exhibit any clear trends as a function of pressure.

To further understand the observed vibrational spectra, we carried out analogous experiments with deuterium. In this case, the interaction potential should remain essentially constant, while the large isotopic mass shift and inverted nuclear spin isomers allow for unambiguous spectral assignments. A comparison of the spectra of the deuterium and protium inclusion compounds in the low-frequency range, shown in Figs. 2(a) and 2(c), allowed us to confirm all the host modes that were predicted by us previously using CRYSTAL09 calculations and also suggest that the calculated E_g and T_2g libration modes of H₂ molecules are related to the observed modes that exhibit very strong pressure dependence (see Fig. S2). While the frequencies of modes dominated by As₄O₆ molecules are predicted very well, the computed absolute frequencies of modes with dominant contributions from H₂ molecules must be shifted by −160 cm⁻¹ and −90 cm⁻¹ for protium and deuterium, respectively, to compare with experimental values, but their evolution with pressure is predicted properly. Interestingly, we were able to obtain the inclusion compound with deuterium at only 1.09(13) GPa (compared with ∼1.5 GPa for hydrogen), which was evidenced by a broad ²H₂ vibron in the inclusion compound clearly separated from the fluid ²H₂ vibron. Attempts to obtain the inclusion compound with protium at similar pressure were not successful, even after 17 h at 1.03(9) GPa.

The ²H₂ vibron in the inclusion compound is also composed of three components (ν₁–ν₃), but at lower frequencies due to the higher reduced mass of deuterium [Fig. 3(c)]. While the absolute frequency difference between these bands is lower than that observed for protium, fits of the ²H₂ spectra recorded at RT with three Lorentzian peaks are always significantly better than with fits using one or two peaks. Moreover, the presence of three vibrons becomes evident at LT when the peaks become sharper [see Figs. 3(d) and S4]. The observed frequencies of the ²H₂ vibrons, both in the inclusion compound and in the fluid/solid, are in very good agreement with the frequencies of the respective ¹H₂ vibrons when scaled by a factor of 1.39, which is close to the theoretical reduced mass ratio of 1.41 (see Fig. S6). At ∼2.5 GPa, the temperature dependence of the relative intensities of the ²H₂ vibrons in the inclusion compound similar to ¹H although the trends for ν₁ and ν₂ are obscured above 200 K due to significant overlap of these peaks [see Figs. 3(d), 4, and S5]. The relative intensity of ν₃ falls from 0.55 at 260 K to 0.07 at 85 K, while the relative intensity of ν₂ grows from RT to 200 K when it reaches ∼0.60 and remains constant down to 85 K. Similar to protium, ν₁ for deuterium grows in intensity as temperature is lowered, while ν₃ decreases in intensity over the same temperature range. We were not able to detect a shoulder on ν₁ in the case of deuterium, but we suspect that this is caused by the very close spacing of peaks, even at 85 K. It is noteworthy that the relative intensities of ν₁ and ν₂ are inverted for deuterium as compared to protium.

Our attempts to determine the LT stability of the inclusion compound at ambient pressure were complicated by irreversible sticking of the DACs that prevented full pressure release at LT. We were, however, able to completely release the pressure during one LT experiment with deuterium, which revealed that the inclusion compound remains (meta)stable at ambient pressure and 85 K for at least 21 h (see Fig. S7 for the recorded Raman spectra).

Noting the observed vibron intensity trends between protium and deuterium and their frequency separations relative to the pure fluid/solid, we decided to carry out an additional HP–LT Raman experiment over a longer time to examine possible ortho–para conversion and make definitive peak assignments. While ortho–para conversion rates vary drastically under pressure,^33,34 conversion in other inclusion compounds is typically slow,^3,4 and we therefore utilized a catalyst to possibly help expedite conversion. Our goal was to study the inclusion compound at the lowest pressure possible to minimize pressure-induced intermolecular coupling, which causes dramatic changes in the relative intensities of the J = 0 and J = 1 vibrons.³⁵ Due to the problems associated with releasing the DAC pressure at LT described above, we were only able to decrease the pressure down to 1.29(5) GPa, which was sufficient to clearly observe the Q₁(1) and Q₁(0) vibrons of solid ¹H₂ [see Fig. 5(d)].

After 66 h at 84(1) K and 1.29(5) GPa, the relative intensities of the solid rotons and vibrons appear to reach constant values, indicating the approach to thermodynamic equilibrium [see Figs. 5(b) and 5(c)]. The relative intensities of the H₂ vibrons for the inclusion compound also reached a plateau albeit within a shorter time span of 42 h. While the intensities of the solid rotons change significantly over time and reflect an ortho concentration moving toward thermal equilibrium, the intensity ratio of the Q₁(1):Q₁(0) fluid vibrons only decreases ∼3% reflecting the stronger vibron scattering for the J = 1 state. Given that the ν₁ and ν₂ relative intensities from the compound change by this same amount and exhibit a similar spacing, we assign the ν₁ and ν₂ vibrons as Q₁(1) and Q₁(0) vibrons of H₂ trapped in the inclusion compound, respectively. This assignment is also supported by the inverse intensity trends observed for deuterium reflecting the difference between ortho and para for the different spin isomers. We note that while the ν₁ and ν₂ relative intensities mirror the ortho and para behavior of the fluid of time, the relative intensity of ν₃ remained constant near 0.15 over the entire period [see Fig. 5(a)].

In order to analyze the Raman spectra that we recorded at LT, it is important to know whether any temperature-induced phase transitions occur upon cooling. We therefore collected powder x-ray diffraction patterns of the As₄O₆·2¹H₂ inclusion compound at 1.6 GPa at LT down to 5 K at the 16-BMD beamline of the Advanced Photon Source. The analysis of the obtained diffraction patterns revealed that there is no phase transition down to this temperature. It is noteworthy that the x-rays used (wavelength of 0.41 Å) scatter mostly from arsenic atoms with negligible diffraction from H₂ molecules. The diffraction patterns rule out a lowering of the symmetry of the host structure formed by As₄O₆ molecules, but do not rule out possible ordering of H₂ molecules in supercells or lowering of the symmetry of the whole structure caused by H₂ molecules (see Fig. S8).

Detailed analysis of H₂ rotons from the inclusion compound is challenging due to the overlap with fluid/solid H₂ present in the cell and overlap in the spectral range where strong As₄O₆ molecular vibrations are present [see Figs. 2(b) and S7(a)]. Nonetheless, it is clear that a low-temperature fine structure is observed for the J = 0 → 2 and J = 1 → 3 manifolds, indicating the lifting of degeneracy due to an anisotropic environment (which was observed for other inclusion compounds, e.g., for hydrogen clathrate hydrates).^3,4 The energy splitting for these bands is quite significant. Partially resolved components for the J = 0 → 2 transitions centered about 360 cm⁻¹ appear to span 40 cm⁻¹, while components appear to span more than 200 cm⁻¹ for the J = 1 → 3 transitions. At least partially hindered rotation, similar to the 2D rotor behavior for hydroquinone clathrate,³⁶ could potentially explain such splitting. While being incapable of capturing the quantum nature of the rotor, molecular dynamics simulations performed on the As₄O₆·2H₂ compound do indeed indicate a non-spherical classical probability distribution of H₂ molecules at the 16c Wyckoff site wherein the distribution is skewed along the crystallographic ⟨111⟩ direction (Fig. S9). Previously, we suggested that H₂ molecules might be aligned along the crystallographic ⟨111⟩ direction based on the calculated energies of an H₂ molecule in different orientations enclosed in an (As₄O₆)₆ cluster simulating the environment of the molecule at the 16c Wyckoff site in As₄O₆·2H₂.² Interestingly, the roton bands appeared much weaker in the spectra of the deuterium inclusion compound, but their presence could be detected in the quenched sample at 85 K [Fig. S7(a)].

Two additional modes were observed at low frequency and appear to be associated with hydrogen (deuterium). For ¹H₂, these bands appear at ∼130 cm⁻¹ and 340 cm⁻¹ at 1.3 GPa and 85 K. As the pressure is increased, they exhibit very strong pressure dependence that agrees with the T_2g and E_g librational modes predicted by DFT computations. It is conceivable that these modes are actually also associated with the J = 0 → 2 manifold, but a shift to 130 cm⁻¹ would represent an unprecedented energy splitting for a molecular inclusion compound. Additional experiments at lower temperature, including inelastic neutron scattering, are needed to fully understand the coupled translational–rotational behavior of hydrogen within As₄O₆·2H₂.

The hydrogen vibrons in As₄O₆·2H₂ are significantly softened from the fluid/solid hydrogen vibrons. From previously characterized H₂ van der Waals compounds, we infer a trend that whenever H₂ forms an inclusion compound with an electron donating substance, the H₂ vibron is red-shifted due to the partial donation of electron density to the H₂ σ^* antibonding orbital. This is the case for the herein studied As₄O₆·2H₂ as well as (H₂S)₂H₂,¹⁴ (H₂Se)₂H₂,³⁷ hydrogen clathrate hydrates,^3,4 (HI)(H₂)₂ and (HI)(H₂)₁₃,³⁸ in which all host molecules contain atoms with stereoactive lone electron pairs that donate some of its electron density to the H₂ molecule. The inclusion compounds with He, Ne, Ar, and Kr exhibit blue-shifted H₂ vibron, indicating repulsive interactions with hydrogen molecules and no charge transfer from noble gas atoms to H₂ molecules.^7,16,17 The effect is reversed in the inclusion compound of hydrogen with xenon where the H₂ vibron is red-shifted, indicating the Xe valence electron density is partially transferred to the H₂ antibonding orbital.^8,9 This is in agreement with the fact that Xe is known for being much more reactive than lighter noble gases and forming numerous chemical compounds. An interesting observation can be made in the series of H₂ van der Waals compounds with CH₄, SiH₄, and GeH₄.^10,12,39 While the H₂ vibron is blue-shifted in the inclusion compound with methane, it is red-shifted for silane and germane. This is caused by the inversion of the X–H bond polarity. The C–H bonds are polarized toward carbon leaving partial positive charge on H atoms making donation of electron density to the H₂ σ^* orbital impossible. From Si on, the X–H bonds are polarized toward H atoms leading to interactions similar to dihydrogen bonds that result in electron density being partially transferred to the H₂ σ^* orbital and weakening the H–H bond.⁴⁰ In the hydrogen–nitrogen van der Waals compounds (N₂)₆(H₂)₇ and (N₂)(H₂)₂, the H₂ vibron is blue-shifted, indicating only repulsive interactions between N₂ and H₂ molecules.^41,42 It is noteworthy that the H₂ vibron is softened when protium or deuterium is dissolved in solid matrices of Ar, Kr, and Xe at ambient pressure and LT (see Ref. 43 and references therein). Calculations revealed that the vibron is softened whenever the H₂ molecule is located in the region of space where the intermolecular interaction potential is attractive.⁴³ Taking this into account, we may conclude that there are two general mechanisms responsible for H₂ vibron softening. One stems from the local intermolecular potential and attraction between host molecules/atoms and H₂ molecules, while the other comes from partial charge transfer between host and H₂ molecules. The latter mechanism dominates repulsive interactions between the host and guest at HP as has been shown for As₄O₆·2H₂.² 

As for the multiple hydrogen vibrons observed for As₄O₆·2H₂, our carefully collected spectra allowed us to conclude that there are three H₂ vibrons from the inclusion compound up to the highest studied pressure. We assign ν₁ and ν₂ as Q₁(1) and Q₁(0) based on (1) the time dependence of their relative intensities at LT indicating slow ortho–para conversion for protium and (2) the change in their relative intensities when going from protium to deuterium reflecting the difference in degeneracy of ortho and para spin isomers between ¹H₂ and ²H₂. As for the shoulder present on ν₁ in LT protium measurements, we attribute the ν_1a main peak and the ν_1b shoulder to A_g and T_2g vibrations, respectively, both of which were predicted by DFT computations and may both exhibit ortho–para contributions in the observed spectra [see dotted lines in Fig. 5(d)].

The origin of ν₃ is more difficult. Our first hypothesis was that H₂ occupies an additional site in the As₄O₆·2H₂ crystal structure. This explanation seems unlikely to be the case since there should be a second pair of vibrons corresponding to ortho and para isomers occupying the second site, which is not the case here [see dashed vertical lines in Fig. 5(d)]. The occupation of a second site is also inconsistent with our MD simulations. We simulated the As₄O₆·2H₂ crystal structure with H₂ molecules initially positioned on the 16c site at RT and multiple pressure points (2 GPa, 4 GPa, and 6 GPa), and we did not observe H₂ molecules migrate to any other sites (see Fig. S9). When H₂ molecules were positioned on the 8a site, they quickly migrate to the 16c site, substantiating its energetic favorability. A possible temperature-dependent occupancy distribution of the 16c site (leading to multiple effective local environments) also seems unlikely based on the absence of ortho–para pairs, the behavior with pressure, and previous single-crystal XRD measurement that give no indication for non-stoichiometry.

It is tempting to assign ν₃ as an additional rovibrational contribution [e.g., Q₁(3)] based on the temperature dependence of the relative intensities of ν₁ and ν₃ both for protium and deuterium (Fig. 4). Here, the depopulation of a higher-energy rotational state is consistent with the observed temperature dependence, as is the apparent transfer of intensity to the next low-lying transition with the same nuclear spin state. However, such an assignment would require an unusual rearrangement of energy levels such that Q₁(3) is observed at a higher frequency than Q₁(0) and Q₁(1). This assignment is also unlikely in that Q₁(2) is never observed and Q₁(3) should not be populated at 80 K [the intensity of solid hydrogen S₀(3) is zero, indicating that the J = 3 rotational state is not occupied in the pure phase].

It is noteworthy that the ν₃ vibron is significantly broader than the ν₁ and ν₂ vibrons, which may suggest a different physical origin. The observed temperature dependence of the ν₃ vibron is consistent with a Boltzmann factor and could indicate possible coupling with a phonon mode. For example, a librational mode or other translational state could potentially couple with symmetric stretching to yield a separate vibron. The final understanding of this vibron will require additional measurements to reveal the detailed energy levels of H₂ trapped within arsenolite. HP solid-state ¹H nuclear magnetic resonance (NMR), lower-temperature Raman, and inelastic neutron scattering measurements are all being pursued to reveal a greater understanding of the dynamics.

See the supplementary material for additional Raman spectra, trajectories from MD simulations, As₄O₆·2H₂ powder diffraction patterns, and tables of calculated Raman frequencies for As₄O₆·2¹H₂ and As₄O₆·2²H₂.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

This work was supported by the Polish National Agency for Academic Exchange. Computations were performed using the Wroclaw Centre for Networking and Supercomputing (http://wcss.pl), Grant No. 260, and the Memex cluster of Carnegie Institution for Science. T.A.S. acknowledges the support from the U.S. Army Research Office under Grant No. W911NF-17-1-0604. Portions of this work were performed at HPCAT (Sector 16), Advanced Photon Source (APS), Argonne National Laboratory. We thank Curtis Kenney-Benson and Dmitry Popov for assistance with cryostat work and x-ray diffraction data collection during beamtime. HPCAT operations are supported by DOE–NNSA’s Office of Experimental Sciences. The Advanced Photon Source is a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357.