Red Bar


We report inelastic neutron scattering results on hydrogen adsorbed onto samples containing single-wall carbon nanotubes. These materials have attracted considerable interest recently due to reports of high density hydrogen storage at room temperature. Inelastic neutron scattering clearly shows the ortho-para conversion of physisorbed hydrogen in a nanotube containing soot loaded with hydrogen. From the rotational J = 0 --> 1 transition, no indication of a significant barrier to quantum rotation is seen.


We have used the neutron scattering facilities at the NIST Center for Neutron Research, to structurally characterize our as-produced SWNT sample using the BT1 powder diffractometer, and then to probe the rotational potential felt by adsorbed hydrogen as a function of temperature with the BT4 filter-analyzer inelastic spectrometer. In the solid, hydrogen behaves as a three-dimensional quantum rotor with energy levels given by
EJ = B J (J+1)
where J is the rotational quantum number and B=7.35 meV is the rotational constant. Here we use the J= 0 --> 1, or para-ortho transition, as a probe of hydrogen binding sites. This transition occurs at an energy close to 2B= 14.7 meV in the pure molecular solid because the molecules are essentially free quantum rotors with no center-of-mass translation. In the gas phase, the continuous spectrum of recoil energies broadens the rotational transition, so the tech- nique is able to discern adsorbed molecules which do not diffuse. Relatively weak interactions between the rotating molecule and its environment will alter this simple energy level scheme; given sufficient spectroscopic resolution, one can deduce accurate information about the adsorbate site symmetry and the nature of the hydrogen-substrate interaction.

This figure shows a similar neutron scattering study for the rotational dynamics of D2 trapped in the octahedral site of solid C60. In this case, the symmetry of the local potential is lowered by the orientational ordering of the C60 molecules, giving rise to a splitting of ~0.7 meV for the ortho-para transition.

A further advantage of neutron scattering in this context is the low absorption in most materials, which facilitates in situ measurements involving cryostats and pressure vessels. These first experiments were performed on as-produced SWNTs that were not purified or processed in a way to produce cutting or opening of tubes, and provide benchmarks for future studies.


Neutron scattering measurements were performed on two different samples; one was obtained from Dr. Heben's group at NREL and the other was purchased from Rice university (tubes@rice). Both samples were annealed under dynamical vacuum for overnight before the hydrogen gas loading. No further processing was performed to purify the samples or to open the tube ends. We obtained almost identical results from both samples and therefore here we report only the results obtained from Rice sample.

For hydrogen loading, the nanotube soot was sealed inside a high pressure aluminum cell, heated to 350 K and placed under dynamic vacuum for ~12 h with a mechanical pump (~10-3 Torr). While the temperature and ultimate pressure may not be sufficient to remove all impurities adsorbed during exposure to the atmosphere, perfect loading conditions may not be necessary to extract information about the location and environment of the adsorbed hydrogen. Hydrogen gas was introduced directly from a standard cylinder at a pressure of ~11 MPa and the system was isolated to monitor for any pressure leaks over a period of 48 h. The experimental setup is shown shematically in Figure 2. To ensure full loading, the sample was cooled in a helium cryostat to 25 K over a period of 18 h while remaining in contact with the hydrogen gas source at 11 MPa. After isolating the hydrogen cylinder, the sample space was evacuated using a mechanical pump for 90 min until evolution of non-adsorbed hydrogen slowed and the vessel pressure was reduced to ~10-3 Torr. Neutron energy loss spectra were recorded on heating between 25 and 65 K, with 10 min equilibration time allowed at each temperature. Based on the measured neutron intensity compared to that from a known quantity of H2 in C60, we estimate the amount of adsorbed hydrogen to be ~0.5 wt% at 25 K.


Figure 3 shows the animation of the neutron energy loss spectra recorded between 10 and 20 meV at several temperatures between 25 and 65 K on warming. The spectra consist of two peaks with intensity ratio of 1:2, corresponding to J=1,M=0 and J=1, M==/-1 states. The splitting of the peaks is about 1 meV, slightly larger than the splitting observed in H2 in C60. The center of the peaks is at ~14.5 meV, only slightly shifted from the value of 14.7 meV characteristic of unhindered rotations.

Figure 3. Temperature dependence of the neutron energy loss spectra for the hydrogen-nanotube system, taking on heating from 25 K and then cooling from 60 K. The spectrum is totally reproducible upon cooling, indicating that hydrogen goes in and comes out from nanotube sample very easily. The dotted lines are the fits to the data assuming two peaks with intensity ratio of 1:2.

Details of the temperature evolution of the J=0 --> 1 transition were followed by fitting the data to a Gaussian peak with a sloping background function. While we found that the peak position and the width were temperature independent, the integrated intensity showed an exponential decrease with increasing temperature, indicating thermally activated desorption of hydrogen from the nanotube surface. This behavior was not observed for H2 in C60, indicating that the binding energy on SWNT surfaces is significantly less than the value of ~11 kJ/mol for H2 adsorbed interstitially. Conversely, the binding energy on SWNT must be somewhat greater than on graphite (4 kJ/mol) since the adsorbed H2 is stable to higher temperatures. We suggest that the relevant site in the present work is the exterior cylindrical SWNT or bundle surface rather than tube interiors (mainly inaccessible in the present sample) or interstitial channels in the rope lattice (probably more similar to octahedral sites in the fullerene solid).

We also attempted to observe the inverse J = 1 --> 0 transition in neutron energy gain with the Fermi-chopper time-of-flight spectrometer at the NCNR, since this would verify our assignment as well as yield information on the rate of ortho-para conversion. However, we were unable to observe a peak that could be ascribed to this transition at any temperature. This is likely due to the extremely rapid conversion of all the ortho hydrogen to the para form via magnetic catalyst particles, resulting in negligible population of the J=1 level.

Finally, it should be noted that no scattering which could be attributed to vibrational motions of the H2 molecules was observed in these measurements. Since one expects to see an in-plane phonon at ~5 meV for a monolayer of H2 adsorbed on graphite, this provides further evidence that hydrogen has not been adsorbed onto graphitic impurities. This was further confirmed by repeating the experiment using a sample of graphite which did not yield any peak in the spectrum as shown in this figure.


The present neutron energy loss results for the J=0 --> 1 transition show that the rotational spectrum of hydrogen is only mildly perturbed by association with the nanotubes, similar to what is observed for solid H2, H2 on graphite, or H2 in C60. At all temperatures the spectrum is splitted by about ~1 meV and it can be fitted to two peaks with intensity ratio of 1:2. This is attributed to the lifting of the degeneracy of the J=1 rotational levels by the adsorption site symmetry, rather than to multiple adsorption sites. The results are reminiscent of those obtained for interstitial hydrogen in C60, where a rotational barrier of ~0.1 kJ/mol was observed. There is an exponential decrease in the intensity of this peak with increasing temperature due to the desorption of the physisorbed hydrogen, indicating a binding energy signi®cantly less than the 19.6 kJ/mol reported for opened SWNTs produced by arc discharge. We conclude that the main binding site under the present loading conditions is the outer surface of isolated tubes and/or tube bundles, with a provisional value for the binding energy of ~6 kJ/mol. These results are valuable as they provide benchmark values to which future hydrogen adsorption experiments on purified and opened SWNTs can be compared.