Direct Evidence for Solid-Like Hydrogen in a Nanoporous Carbon Hydrogen Storage Material at Supercritical Temperatures

: Here we report direct physical evidence that confinement of molecular hydrogen (H 2 ) in an optimized nanoporous carbon results in accumulation of hydrogen with characteristics commensurate with solid H 2 at temperatures up to 67 K above the liquid-­‐vapour critical temperature of bulk H 2 . This extreme densification is attributed to confine-­‐ ment of H 2 molecules in the optimally-­‐sized micropores, and occurs at pressures as low as 0.02 MPa. The quantities of contained, solid-­‐like H 2 increased with pressure and were directly evaluated using in-­‐situ inelastic neutron scattering and confirmed by analysis of gas sorption isotherms. The demonstration of the existence of solid-­‐like H 2 challenges the existing assumption that supercritical hydrogen confined in nanopores has an upper limit of liquid H 2 density. Thus, this insight offers opportunities for the development of more accurate models for the evaluation and design of nanoporous materials for high capacity adsorptive hydrogen storage.

hydrogen exists as a gas, and will not form a higher--density bulk liquid or a solid, except under extreme conditions of high pressure (e.g., > 5 GPa). 2,3 Subsequently, the efficient and economic storage of molecular H 2 remains a major tech-nological challenge. 4,5 One option for increasing storage densities is via adsorption of H 2 into microporous materials, that is, materials with pore diameters < 2 nm 6 . In such mate-rials, densification of H 2 is promoted via the enhancement of the attractive van der Waals interactions between adsorbed H 2 molecules and the solid substrate, arising from overlap-ping potentials from opposite pore walls. Evaluation of gas storage capacities of promising nanoporous materials gener-ally involves measurement of the Gibbs excess uptake via isothermal gas sorption, with the subsequent conversion of the excess to absolute H 2 uptake requiring an estimate of the adsorbed hydrogen density. 7 As the density of H 2 inside the micropores is difficult to probe experimentally, the maxi-mum (limiting) density of H 2 is generally approximated to be the same as liquid hydrogen (i.e., a uniform density of 77 kg m --3 at the triple point), despite temperatures of adsorptive storage typically exceeding the bulk critical temperature. 8 One of the few experimental techniques that can directly access information on the state of the adsorbed H 2 inside a porous material and potentially validate the assumption of liquid--like adsorbed phase densities is neutron scattering. Neutrons are highly sensitive to 1 H due to its large incoher-ent neutron scattering cross section. While there have been numerous neutron diffraction studies investigating the bind-ing of hydrogen to strong adsorption sites in metal--organic frameworks (MOFs), 9,10,11 due to the magnitude of the inco-herent scattering background, 1 H 2 is almost always substitut-ed by D 2 . The differences in molecular weights may intro-duce isotope effects that will affect bond distances, vibra-tional energies and packing densities. To avoid the need for deuteration and to allow investigation of non--crystalline materials, in this study we used inelastic neutron scattering (INS), a technique that is not hampered by the 1 H incoherent scattering and which measures vibrational motions and thus the binding strength of atoms and molecules. While INS measurements are typically performed at temper-atures below 25 K to maximize resolution of the vibrational spectra, 12,13 here we combined INS measurements with volu-metric gas sorption experiments to probe the phase behav-iour of supercritical hydrogen at 77 K (a temperature that is more practically relevant for H 2 storage applications) in a nanoporous carbon material and show direct physical evi-dence for an accumulation of solid--like H 2 in the pores. The INS measurements were only possible due to modifica-tions to the TOSCA instrument at the ISIS neutron facility, which enabled measurement at high resolution (ΔE / E < 1.25 %, where E is the energy lost by the incoming neutron) over the widest range of energy transfer of any INS instrument in the world. The improved high resolution at low energies allowed quantitative analysis of the elastic re-gion of a scattering spectrum (where little or no energy is transferred between the incident neutron and the target H 2 molecule) as a function of gas pressure, with simultaneous monitoring of the inelastic regions, to provide information on the state (gaseous, liquid or solid) of the H 2 in the pores. INS spectra were collected on H 2 dosed onto a standard reference material of TE7 activated carbon beads (from MAST Carbon International, UK) at eight gas pressures (0.016,0.070, 0.160, 0.301, 0.630, 0.998, 2.070, and 3.500 MPa), with ~12 h data collection periods. This material was selected as it presents a reasonably chemically--homogenous adsorb-ing surface, allowing unambiguous analysis of INS spectra, and has a modal nanopore diameter within the pore size range for maximizing interactions with H 2 14 (~0.7 nm as determined via N 2 sorption at 77 K - see Figure S1 in Supple-mentary Information). The chemical and structural proper-ties of the TE7 carbon were thoroughly characterized 15 and the excess H 2 sorption isotherm for this material was meas-ured at 77 K to a maximum pressure of 17 MPa (see Figure S2 in Supplementary Information). For the INS experiment, a room--temperature equilibrium mixture of para--H 2 (J = 0, where J is the rotational quantum number) and ortho--H 2 (J = 1) was used for the in--situ gas dosing at 77 K (see Figure  1a). The use of so--called "normal hydrogen", which includes the para and ortho nuclear spin isomers of molecular hydrogen resulted in distinct characteristic features in both the elastic and inelastic regions of the neutron scattering spectrum. Due to the paramagnetic nature of activated carbons, the statisti-cal population of para--H 2 : ortho--H 2 at 77 K rapidly equili-brated to a 1:1 mixture in the sample. 16

RESULTS AND DISCUSSION
The INS energy loss spectra of the H 2 in the pores at 77 K ( Figure 1b) showed two important and remarkable features at each H 2 loading pressure measured. First, the intense, sharp peak at ~0 meV due to elastic scattering by ortho--H 2 indicat-ed the presence of dense liquid--or solid--like hydrogen. While both condensed phases will show a sharp elastic scat-tering peak due to far higher densities of ortho--H 2 molecules than in the gas phase, the elastic peak from the more mobile liquid phase will typically be broadened relative to the same peak from solid H 2 , as a result of quasielastic interactions. The full--width at half--maximum (FWHM) of the elastic peak of 0.3 meV at the lowest adsorption pressure measured here (0.016 MPa H 2 ) remained approximately equal to TOSCA's instrumental peak resolution (indicated by the horizontal bar of width 0.3 meV in Figure 1c), taking into account the significant contributions from multiple--scattering and self-shielding from the >10 cm 3 sample size, suggesting that the H 2 contributing to this peak had limited mobility. A full analysis of the FWHM of the elastic line is provided in the Supplementary Information (see Figures S3 and S4). The second prominent feature present in all of the spectra was a well--resolved peak at ~14.7 meV (Figure 1d), which is only present for the para--to--ortho transition in immobilized H 2 . 12,13,17,18 This peak (commonly referred to as the "rotor line", as it corresponds to the free, unperturbed rotation of molecular H 2 ) denotes the presence of H 2 that is strongly pinned in all three dimensions and lacks translational free-dom to recoil. It is not present in the INS spectra for either bulk liquid or gaseous para--H 2 (see Figure S5 in Supplemen-tary Information), and indicates a population of solid--like H 2 which systematically increases with increasing H 2 loading pressure. A full analysis of the FWHM of the rotor line is provided in the Supplementary Information (see Figure S6 and S7). The 14.7 meV rotor line was also confirmed to also be present in the INS spectrum of the same material dosed with H 2 at a temperature of 100 K (see Figure S8 in Supple-mentary Information). As the total neutron scattering intensity in each INS spec-trum was normalized to the number of neutron counts, the integrated intensities under the peaks over different regions of the spectrum could be evaluated to determine relative amounts of H 2 in different states as a function of H 2 dosing pressure ( Figure 2). The total integrated intensity over the entire inelastic range (2--500 meV) was dominated by the contribution from gaseous H 2 in and around the sample and increased linearly with pressure ( Figure 2c), as would be expected from the gas equation of state. 1 The area under the elastic peak, determined by numerical integration from --2 meV to +2 meV after subtraction of a linear background and indicating the amount of H 2 in a strongly bound con-densed state, increased rapidly at low pressures before start-ing to plateau at pressures >1 MPa (Figure 2d). The integrat-ed intensity under the 14.7 meV rotor line, determined using a Gaussian peak shape after subtraction of a fourth--order polynomial background, corresponded solely to the amount of solid--like H 2 in the material. The presence of the 14.7 meV rotor line even at the very lowest pressures of H 2 studied (0.016 MPa) suggests direct accumulation of H 2 with solid-like characteristics, rather than a gaseous phase that increas-es in density with increasing pressure. Figures 2d and 2e show the dependence of the integrated intensity of the rotor line with pressure matches that of the elastic peak within the uncertainty margins resulting from the lower count statistics of the lower--intensity rotor line. The closely mirrored pro-files of the integrated intensities of the two independent peaks suggested strongly that the densification of both the ortho--and para--H 2 were a result of the same phenomenon and were due to the presence of the solid--like hydrogen.
The pressure dependence of the H 2 accumulation measured by INS was compared to amounts measured by volumetric gas sorption using a Sieverts--type apparatus at 77 K ( Figure 3). The integrated intensity over the elastic region of the spectrum (chosen for comparison due to the superior count statistics) in arbitrary units was scaled to result in a least--squares best fit to the calculated weight percent of H 2 in the adsorbed phase from volumetric gas sorption meas-urements. The scaled INS integrated intensities were found to be strongly correlated to the calculated total amount of adsorbed H 2 expressed in weight percent relative to the dry, evacuated carbon sample (see Figure 3 and Tables S1 and S2 in Supplementary  Information), with the concurrence of the onset of the plateau region in the INS intensities and the gas sorption data indicating that the accumulation of the solid hydrogen has an effective upper limit, which may signify the point at which the nanopores are completely filled with adsorbed H 2 . The original and scaled data are in Tables S1 and S2 in Supplementary Information, along with details of the modelling.
Theoretical predictions of the density of the adsorbed H 2 (termed the "adsorbate") in the pores, using a development of earlier analysis 19,20 applied to the experimental high-pressure Gibbs excess isotherm (to 17 MPa) for H 2 adsorbed on the TE7 carbon beads at 77 K, was also consistent with the evidence from INS for the formation of solid--like H 2 . The excess data was modelled using the Tóth equation 21 for ad-sorbate filling of the pore space, to yield an estimate of the density of the adsorbed H 2 phase (assumed constant) of 101±2 kg m --3 (see Figure S2 in Supplementary Information). This is significantly higher than the maximum density of liquid H 2 , 77 kg m --3 at the liquid--solid--vapour triple point (13.96 K and 0.00736 MPa) 1 and is closer to the density of bulk solid H 2 , which is a highly compressible solid with den-sities >87 kg m -3 . 16 The powerful combination of INS measurements and gas sorption analysis points to a bulk densification phenomenon that, while consistent with modelling (see, for example, the prediction of a "solid--like phase with densities higher than bulk solid hydrogen" from Dundar et al.'s modelling of su-percritical hydrogen sorption on MOFs) 22 had not been pre-viously observed experimentally. Past experimental observa-tions consistent with localized regions of solid--like densities of adsorbed H 2 at elevated temperatures have generally been attributed to the strength of specific adsorption sites in MOFs and zeolites. For example, refinements of neutron diffraction data at 4 K indicate that some crystalline MOF materials support D 2 --D 2 intermolecular separation distances that are shorter than the 0.36 nm found in solid H 2 . 10,23--25 In these studies, the local surface densification of H 2 was as-cribed to the strong interactions between the H 2 and the unsaturated metal centres in the frameworks. Similarly, H 2 rotor lines previously reported in INS spectra of zeolite samples at temperatures up to 70 K show a shift in energy, due to the influence of the strong binding of H 2 on specific adsorption sites. 18,26 Short range order and liquid--like behav-iour of D 2 has also been predicted to exist in areas between metal sites in zeolites and MOFs at 50 K, to explain broaden-ing effects in neutron diffraction patterns. 11 Carbon surfaces, however, are known to only have very weak interactions with H 2 . The phenolic resin--derived TE7 carbons, in particular, have been shown via temperature programmed desorption to have only small proportions of surface oxygen groups, 27,28 meaning that they are likely to have limited surface func-tionality. This indicates that the pseudo--condensation of supercritical gas seen here at 77 K is instead due to confine-ment effects in optimally--sized nanopores. 14 Comparative, single pressure (1 bar) H 2 dosing INS measure-ments on onion--like carbon nanomaterials (OLC--1750) hav-ing negligible proportions of nanopores less than 10 Å in diameter 29,30 show no such peak in the 14.7 meV region, supporting the hypothesis that the pore dimension is a criti-cal factor. Similarly, while other activated carbon materials (TE7--20 and TE3 from MAST carbon) and AX--21 (Anderson Development Co) show small rotor peak contributions (see Supplementary Figure S9), the integrated intensity is not proportional to the total micropore volume, indicating that only a fraction of the micropores contribute to this effect and that a very narrow pore size distribution is required. This has been shown experimentally in the case of carbide--derived carbons, 31 for which higher than liquid H 2 densities were calculated from 77 K sorption isotherms for ~0.6 nm diame-ter pores. 32 Molecular simulation and modelling of hydrogen in carbon nanomaterials seems to support the possibility of densifica-tion of adsorbed hydrogen to greater than liquid densities. For example, there have been estimates of adsorptive capaci-ties from Grand Canonical Monte Carlo molecular simula-tions (with quantum effects estimated using the Feynman--Hibbs effective potential) and density functional theory (without the dispersion correction) that predicted elevated levels of densification for H 2 in 0.3 nm carbon slit pores and carbon nanotubes, 33,34 equating to densities in the region of 80 kg m -3 at 0.1 MPa and 77 K, while theoretical modelling of supercritical adsorption isotherms has predicted a transition to a solid--like phase of H 2 in activated carbon at 40 K. 35 Ex-perimental room--temperature small--angle neutron studies of 0.9 nm pores in carbon have also estimated greater than liquid densities of H 2 at pressures of 20 MPa. 36 Confinement effects are known to induce shifts in the phase transition temperatures for sub--critical adsorbed phases 37 and, thus, the pseudo--condensation of supercritical gas re-ported here could be a general phenomenon. It is, therefore, possible that the presence of the ~14.7 meV rotor line in INS studies of other highly nanoporous carbons at supercritical temperatures 38,39 is likely to be similarly indicative of the presence of densified solid--like H 2 (Supplementary Infor-mation, Figure S9). This is, however, the first time that the observation of this phenomenon has been correlated to gas sorption measurements and shown to be consistent with accumulation of solid--like H 2 with pressure.

CONCLUSION
Experimental evidence for a solid--like H 2 in the pores of TE7 carbon at 77 K clearly demonstrates the potential for further development of adsorptive hydrogen storage materials con-taining micropores of an optimum size. Due to its relatively low micropore volume (< 0.5 cm 3 g -1 , see Supplementary Information) the 3 wt% H 2 uptake of the TE7 material at 77 K and 17 MPa (which is ~80 % of the estimated maximum up-take at this temperature) remains low compared with the current US Department of Energy targets for onboard H 2 storage systems for vehicles 40 (7.5 wt% of useable H 2, based on the weight of the entire hydrogen storage system, includ-ing the tank, cooling and delivery systems). However, a ma-terial with a much higher, and not unreasonable, micropore volume of 1.5 cm 3 g -1 which is 80 % filled with H 2 adsorbate of density 100 kg m -3 , would display an impressive uptake of 12 wt%. This compares favourably with the high observed H 2 uptake of 10 wt% in MOF NU--100 at 5.6 MPa and 77 K. 41 This work also counters the widely--held assumption of an upper limit of liquid--like adsorbed phase density, and, thus, will aid in advancing models for the evaluation of gas storage capaci-ties in nanoporous materials, and could lead to the design of higher--capacity nanoporous adsorbents.

METHODS
The reference sample of TE7 activated carbon beads (sourced from MAST Carbon International, UK) was produced from a carbonized phenolic resin--based material activated at high temperature (750 °C) in a steam atmosphere. The BET nitro-gen specific surface area at 77 K was measured to be 960±50 m 2 g -1 obtained from low pressure (up to 0.1 MPa) nitrogen sorption measurements at 77 K with a 60 min equi-libration time. 15 The micropore volume, evaluated from the Dubinin--Radushkevich method is 0.43±0.03 cm 3 g --1 . The skeletal density of the sample was measured using a He pycnometer (Micromeritics Ac--cuPyc1330) and was established as being 1.90±0.03 g cm --3 .
The OLC--1750 carbon onions where synthesized by vacuum annealing of detonation nanodiamond powder (Ray Tech--nologies Ltd., Israel). Placed in a graphite crucible, the nanodiamond powder was annealed in vacuum (10 --4 to 10 --5 mbar) for 3 h at 1750 °C in a water--cooled high temperature furnace with tungsten heaters (Thermal Technology, USA). 29,30 High--pressure (up to 20 MPa) hydrogen adsorp-tion/desorption measurements were performed on a Hiden HTP--1 Sieverts--type volumetric gas sorption analyser with ultra--high purity (Air Products BIP--Plus, 99.99996 %) hy-drogen at 77 K using a liquid nitrogen bath for temperature control. Prior to hydrogen uptake measurements the 150 mg samples were degassed at 623 K for 8 h under a vacuum of 0.1 mPa prior to each isothermal measurement in order to remove moisture and desorbed gases from the surface. All isotherms were fully reversible and repeat isotherms for different samples were reproducible to within 0.3 % of meas-ured amounts adsorbed.
The INS spectra were collected on the TOSCA inelastic neu-tron scattering beamline at the Rutherford Appleton labora-tories in the UK, which has an energy window from --3 meV to +500 meV. The full--width at half--maximum instrumental resolution is 300 µeV over the range --3 meV to +3 meV (i. e., in the elastic region) and in the range 3 to 500 meV the ener-gy resolution is ΔE / E < 1.25 %. A 10 g sample of carbon was degassed via heating ex--situ at 623 K for 8 h under high vac-uum (0.1 mPa), then loaded in an Ar glovebox into a high pressure (7 MPa) stainless steel sample can. Temperature control was supplied by a standard cryofurnace ancillary. Normal hydrogen gas (Air Liquide, 99.999 % purity) was dosed into the sample and thermally equilibrated at 77 K before the pressure was recorded using a baratron and a high--pressure transducer. The data were corrected for the presence of terminal H atoms in the carbon by subtraction of 12 h background scans of the degassed sample under dynam-ic vacuum at the measurement temperature (77 K). Note that the spectra for the degassed carbon sample did not show a peak at 14.7 meV. Data were accumulated for 700 µA h, with up to three spectra being collected at each pressure over collection periods of 8--12 h (pressures = 0.016, 0.070, 0.160, 0.301, 0.630, 0.998, 2.070, and 3.500 MPa). The data pro-cessing and peak integration was performed using the Man-tid software (available from http://www.mantidproject.org).

ASSOCIATED CONTENT Supporting Information
Details of the materials characterization, experimental methods, raw data, analysis and modelling are available free of charge via the Internet at http://pubs.acs.org.

Corresponding Author
V.Ting@bath.ac.uk