Full text of DOI:10.1557/JMR.2002.0326

ARTICLES
Large cryogenic storage of hydrogen in carbon nanotubes at
low pressures
B.K. Pradhan,a) A.R. Harutyunyan, and D. Stojkovic
Department of Physics, The Pennsylvania State University, University Park, Pennsylvania 16802
J.C. Grossman
Lawrence Livermore National Laboratory, 7000 East Avenue, Livermore, California 94550
P. Zhang, M.W. Cole, and V. Crespi
Department of Physics, The Pennsylvania State University, University Park, Pennsylvania 16802
H. Goto and J. Fujiwara
Honda R&D Co., Ltd., Wako Research Center, Saitama 351-0193, Japan
P.C. Eklundb)
Department of Physics, The Pennsylvania State University, University Park, Pennsylvania 16802
(Received 4 January 2002; accepted 15 May 2002)
We report up to 6 wt% storage of H2 at 2 atm and T ס 77 K in processed bundles of
single-walled carbon nanotubes. The hydrogen storage isotherms are completely
reversible; D2 isotherms confirmed this anomalous low-pressure adsorption and also
revealed the effects of quantum mechanical zero point motion. We propose that our
postsynthesis treatment of the sample improves access for hydrogen to the central
pores within individual nanotubes and may also create a roughened tube surface with
an increased binding energy for hydrogen. Such an enhancement may be needed to
understand the strong adsorption at low pressure. We obtained an experimental
isosteric heat qst ס 125 5 meV. Calculations are also presented that indicate disorder
in the tube wall enhances the binding energy of H2.
I. INTRODUCTION
materials have been developed with very high experi-
mental specific surface areas of As ∼ 1000 to 3000 m2/g.3
These materials tend to be disordered, with convoluted
surfaces and predominately sp2 C–C bonding. Despite
these impressively large surface areas, these materials
have not shown promise for hydrogen storage.2,4
Recently, significant wt% H2 storage has been reported
in nanofilamentary carbon at room temperature:
5–10 wt% in bundles of single-walled carbon nanotubes
(SWNT);5,6 4–8 wt% in larger diameter SWNTs;7
10–20 wt% in alkali-metal-doped SWNTs;8 0.4–4 wt%
in electrochemically H2-loaded SWNTs;9–11 approxi-
mately 50 wt% in carbon nanofibers.12 Cryogenic H2
storage of approximately 8 wt% in SWNTs has been re-
ported at high pressures around 100 atm.13 Unfortu-
nately, none of the claims of hydrogen storage exceeding
the DOE limit have been confirmed in other laboratories.
In some cases, counter claims asserting experimental er-
ror have been published.14–17 For recent reviews of H2
storage in solids, see Refs. 18 and 19. The unsettled ex-
perimental situation suggests that a systematic character-
ization of the carbon substrate may provide crucial
insights. In this study, we show that the wt% hydrogen
For at least a decade, fuel-cell-powered ground trans-
portation has been recognized as a much-needed tech-
nology to ameliorate environmental problems associated
with hydrocarbon combustion.1 Currently, the technol-
ogy for powering vehicles (i.e., fuel cells) is significantly
more advanced than that for storing the fuel (hydrogen).
One figure of merit for hydrogen storage is the weight
percent (wt%) hydrogen stored relative to the weight of
the storage medium. The United States Department
of Energy has recently estimated that approximately
6–7 wt% storage is a benchmark for a viable ground
transportation technology.2 Hydrogen storage in a light-
weight carbon material would therefore be particularly
attractive.
The physical adsorption of gases within the micro-
pores (diameter <2 nm) of carbon materials has been ac-
tively studied for more than 50 years.3 Many such carbon
a)Present address: Columbia Chemicals Co., 1800 West Oak Com-
mons Court, Marietta, GA 30062.
b)Address all correspondence to this author.
e-mail: pec3@psu.edu
J. Mater. Res., Vol. 17, No. 9, Sep 2002
© 2002 Materials Research Society
2209
http://journals.cambridge.org
Downloaded: 11 Mar 2015
IP address: 169.230.243.252

B.K. Pradhan et al.: Large cryogenic storage of hydrogen in carbon nanotubes at low pressures
adsorption sensitively depends on the postsynthesis treat-
are conducted at room temperature and below, any metal
in the sample can only decrease the wt% storage, as the
metal cannot store hydrogen (as a hydride) at these tem-
peratures. In Figs. 1(a)–(d), we display HRTEM images
for a sample of SWNT as-received [Fig. 1(a)] and
at successive processing steps [Figs. 1(b)–1(d)] (cf.
Table I); H2 storage was studied at each of these stages.
For the as-received material, HRTEM images show pri-
marily bundles of nanotubes and carbon-coated Ni–Y
catalyst particles. To remove the undesirable amorphous
carbon that coats the SWNT bundles, we first selectively
oxidized the amorphous carbon at T ∼ 350 °C in flowing
dry air for 30 min. However, it is difficult to remove
nanocrystalline graphite particles present in the soot.
This oxidation also weakens the ordered sp2 carbon coat-
ing that passivates most of the metal catalyst particles in
the soot so that 75% to 90% of the metal can be removed
via a reflux in a mild mineral acid, i.e., 4.0 M HCl at
T ס 130 °C for 18 h. After the samples were refluxed
ment of the materials, which is revealed via careful char-
acterization with high-resolution transmission electron
microscopy (HRTEM), Raman scattering, and adsorp-
tion isotherms of N2 Brunauer–Emmett–Teller (BET),
H2, and D2.
II. EXPERIMENTAL
The arc-derived SWNT material used in the present
studies was obtained from CarboLex, Inc.20 All data re-
ported here are collected using material from the same
batch of SWNT material. Temperature-programmed-
oxidation (TPO) revealed that the CarboLex material
contained approximately 6 at.% residual metal catalyst.
Most of this residual metal can be removed, as described
below. This purification is important because the storage
performance of the carbon sample is based on wt% up-
take of hydrogen. Since our hydrogen storage experiments
FIG. 1. Room-temperature transmission electron microscopy images of arc-derived carbon nanotubes at various stages of postsynthesis proc-
essing: (a) as-received; (b) after selective oxidation/mild HCl reflux; (c) after selective oxidation followed by aggressive HNO3 reflux; (d) after
step (c) followed by a heat treatment at 1000 °C in vacuum.
2210
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 11 Mar 2015
IP address: 169.230.243.252

B.K. Pradhan et al.: Large cryogenic storage of hydrogen in carbon nanotubes at low pressures
in HCl (or HNO3), we carefully washed the sample in
The spectra were taken at 1064 nm with a Nd:YAG laser
that in our tube diameter range (approximately 1.0–
1.6 nm) excites only the semiconducting tubes via reso-
nant Raman scattering.21 The dominant spectral features
include the low-frequency radial breathing modes in the
range approximately 150 to 200 cm−1 and the higher fre-
quency tangential modes in the range 1500 to 1600 cm−1
[Fig. 1(a)]. The Raman-active radial breathing mode is
the clear spectroscopic signature of a cylindrical seam-
less SWNT; the diameter dependence of the mode fre-
quency for tubes in the range 1.0 < d < 1.6 nm is
deionized water to remove residual acid. On the basis of
Raman scattering spectra and HRTEM images, it appears
that HCl does not react with the SWNT (but HNO3
does—see below). By “react” we mean “cause damage”
to the nanotube wall. A typical HRTEM image after a
selective oxidation/HCl treatment [Fig. 1(b)] shows sig-
nificant reduction in the catalyst residue (Ni–Y) content
(i.e., to approximately 1.5 at.% metal). The amorphous
carbon that coats the bundle exterior has also been
largely removed. A second sample was selectively oxi-
dized in the same way and then exposed to a considerably
more aggressive oxidative treatment in refluxing 2.6 M
HNO3 at T ס 130 °C for 28 h. Although the HNO3
reflux removed almost all of the residual catalyst from
the SWNT materials (i.e., less than 0.2 at.% metal re-
mained), significant damage to the tube wall (i.e., small
holes) can be seen in the HRTEM image [Fig. 1(c)]. This
extensive wall damage can be largely reversed by heating
the material to 1000 °C in a high vacuum (P ∼ 10−8 torr)
for 20 h [Fig. 1(d)]. The apparent damage in our sample
is considerably higher than reported previously. We ex-
amined samples that had received this HNO3 reflux and
that were then vacuum annealed at high T (1000 °C,
20 h). Many of the functional groups attached to the
tube ends and at wall defects are expected to be desorbed
in the 1000 °C vacuum treatment. During 1000 °C an-
neals in our thermogravimetric analyzer (TGA) appara-
tus, we observed sample weight losses in the range
5–45%, depending on the sample and its chemical
treatment.
approximately ␻ ס (224 cm−1 nm)/d (nm) ⌬, where
R
⌬ ∼ 12 cm−1 is a correction term for the tube–tube inter-
action within a bundle.21 The radial mode frequency and
line width are particularly sensitive to C-atom disorder in
the tube wall and also to changes in the tube “bundling”
III. RESULTS AND DISCUSSION
A. Raman scattering studies
FIG. 2. Room-temperature Raman spectra of bundles of arc-derived
carbon nanotubes at various stages of post synthesis processing: (a)
as-received; (b) after selective oxidation/mild HCl reflux; (c) after
selective oxidation followed by aggressive HNO3 reflux; (d) after step
(c) followed by a heat treatment at 1000 °C in vacuum. The Raman
spectra were taken using 1064-nm excitation.
Raman scattering is a sensitive probe of the structure
and bonding in carbon materials, particularly carbon
nanotubes.21–23 Figures 2(a)–2(d) show Raman scatter-
ing spectra for these four sets of samples (cf., Table I)
taken at room temperature in the range 100 to 1700 cm−1.
TABLE I. Sample history and hydrogen storage at 77 K and 1 atm.
Vac. anneal
T (°C)/time (h)
Sample
Sample history
SSAd (m2/g)
H2 (wt%)
Metal (at.%)e
A
B
C
D
E
As-prepared SWNTs
SO,a HClb
270
320
470
180
250
0.52
1.1
2.0
0.32
6.4
250/12
250/12
1000/20
250/12
6
1.5
1.5
<0.2
<0.2
SO,a HClb
SO,a HNO3
c
c
SO,a HNO3
1000/20
a(SO) Selective oxidation at 350 °C for 30–45 min in flowing air (100 sccm).
bReflux in 4 M HCl at 120 °C for 6 h.
cReflux in 2.6 M HNO3 at 130 °C for 30 h.
dSpecific surface area (BET).27
eDetermined by temperature programmed oxidation (TPO).
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2211
http://journals.cambridge.org
Downloaded: 11 Mar 2015
IP address: 169.230.243.252

B.K. Pradhan et al.: Large cryogenic storage of hydrogen in carbon nanotubes at low pressures
brought about by postsynthesis treatment (i.e., via ⌬).
reflux should still persist and broaden the Raman line-
widths and contribute to the D-band intensity. These
functional groups should attach themselves on the edges
of holes in the tube wall. The disappearance of the radial
band can be attributed to the disorder in the tube wall and
is identified with the softening and broadening of the
Raman-active radial mode.
The SWNT tangential displacement modes observed
near approximately 1600 cm−1 are related to the high-
frequency vibrational modes of a flat graphene sheet.
They are also sensitive to disorder in the tube wall and
less sensitive to tube–tube interactions and bundling or
debundling.
The spectrum (a) shown in Fig. 2 is typical of arc-
derived material made with Ni–Y catalyst:24 the broad
band centered near approximately 1300 cm−1 is weak
relative to the band at approximately 1600 cm−1
[Fig. 2(a)]. This band (at 1300 cm−1), commonly called
the D band, has been observed in many sp2-bonded car-
bon materials and is associated with disorder in the hex-
agonal carbon network. In the case of a multicarbon
phase material containing SWNTs, the D band could
arise from either the sp2 carbon present as nanoparticles,
as a thin amorphous coating on the nanotube bundle ex-
terior, or from disorder within the hexagonal carbon net-
work of the nanotube wall itself. The fact that spectrum
(a) has a weak D band indicates that the sample contains
a small amount of disordered sp2 carbons and that the
nanotube walls have a low number of defects. Chemical
processing (particularly, in the case of HNO3), as shown
below, was found to have a dramatic effect on the dis-
order in the system as seen via the D band. Comparing
spectrum (a) to that of spectrum (b) in Fig. 2, we see that
selective oxidation, followed by HCl reflux (Table I), has
no significant effect on the Raman spectrum of the
SWNT material. This was also verified by a careful
Lorentzian line-shape analysis for the full width at half-
maximum (FWHM) of the Raman bands. No change in
the FWHM following selective oxidation and HCl reflux
was observed.
The final high-T vacuum anneal at 1000 °C (approxi-
mately 20 h) almost completely restores the Raman
spectrum [spectrum (d), Fig. 2] to that of the as-prepared
(starting) material [spectrum (a)]: The D band is
significantly reduced in intensity, the tangential bands
have narrowed considerably, and a narrow radial
band is regenerated. As proved by the Raman spectrum
the restoration of structural order to the SWNT is
remarkable. Apparently, SWNTs can “heal” at signifi-
cantly lower temperatures than needed for graphitization
(i.e., T > 2000 °C). A careful line-shape analysis of
the Raman bands after the vacuum anneal does, how-
ever, reveal a residual line broadening associated with
remnant disorder in the SWNT wall. A pervasive rem-
nant disorder might enhance the binding energy of H2, as
discussed later.
B. Adsorption studies
Hydrogen storage was studied on 75-mg samples at
temperatures between 77, 87, and 300 K and at pressures
between 0 and 20 bar. Data were collected gravimetri-
cally, using a high-pressure TGA (model IGA-3, Hiden,
Inc., UK) using ultrahigh-purity (UHP) gases (H2, D2)
(99.999%) that were passed through an oxygen/moisture
trap (MegaSorb Gas Purifier, Supelco Inc., PA). The
equilibrium adsorption data (i.e., weight enhancement)
at each pressure were collected after waiting for suffi-
cient time for the sample mass to reach an equilibrium
value at the particular pressure (this was usually achieved
in less than 10 min). The observed mass uptake was
corrected for buoyancy.25 Just prior to the hydrogen
sorption study, the samples were first heated in situ at a
vacuum of 10−8 torr for a temperature/time indicated
in Table I.
The postsynthesis processing for each sample is also
summarized in Table I, together with the atomic percent
metal residue and the specific surface area (SSA) (m2/g).
The vacuum heat-treatment and the SSA measurements
(N2 isotherms at T ס 77 K) were made in situ in the
TGA just prior to H2 loading. Our experimental SSA
values (180–470 m2/g) are quite low compared to the
maximum geometric surface area (approximately
1350 m2/g) expected for bundled tubes with mean diam-
eter 1.4 nm. We arrive at this theoretical SSA in the
following way. We assume large bundles containing hun-
dreds of tubes. These tubes are all open, and we neglect
the area on the outside (external surface) of the bundle, as
However after selective oxidation of the as-prepared
material, refluxing in the much more aggressive HNO3
for 28 h was found to alter the Raman spectrum dramati-
cally [Fig. 2(c)]: the D band at approximately 1300 cm−1
is much stronger. The tangential band broadens signifi-
cantly. {In disordered sp2 carbons the width of the
1600 cm−1 band (or G band) is also used as a measure of
disorder [for detail, see M.S. Dresselhaus, G. Dressel-
haus, M. Pimenta, and P.C. Eklund, Analytical Applica-
tion of Raman Spectroscopy, edited by M. Pelletier
(Blackwell Science Ltd., Oxford, UK, 1999), p. 187.]}
The radial band broadens to the point of near invisibility.
We conclude that the Raman spectrum clearly indicates
that the HNO3 reflux has damaged the SWNT, consistent
with the TEM images. After this HNO3 reflux (followed
by washing in deionized water), we heated the sample at
10−6 torr vacuum to approximately 250 °C for 12 h. This
temperature should be sufficient to remove any residual
+
HNO3 as NOx, so the Cp NO3 intercalation compound
has been essentially destroyed. However, any function-
alization of the tube wall at defect sides caused by HNO3
2212
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 11 Mar 2015
IP address: 169.230.243.252

B.K. Pradhan et al.: Large cryogenic storage of hydrogen in carbon nanotubes at low pressures
well as the small area associated with the interstitial
of Fig. 3 is the very low H2 (1–2 atm) overpressure
required for significant storage in the processed SWNT
material. In our data, the wt% H2 stored is referenced to
the weight of the sample; no correction to the wt% data
has been made for residual metal or other carbon phases.
Furthermore, the adsorption/desorption data are com-
pletely reversible. This reversibility rules out the adsorp-
tion at 77 K of impurity gases such as H2O and O2. We
can compare our data for the HNO3-treated material to
that of Ye et al.,13 who reported a maximum of approxi-
mately 8 wt% storage at T ס 77 and approximately
100 atm. Their H2 over-pressures are a factor of 20–40
higher than required in our material to achieve similar
storage. This remarkable change in the pressure scale
suggests to us that the structural modifications we have
created in the individual SWNTs and tube bundling may
play an important role in enhancing the binding energy
for H2 adsorption.
The inset to Fig. 3 compares the T ס 77 K isotherms
for hydrogen and deuterium (D2) on the same sample.
Data were first collected with H2. Next the sample was
degassed in vacuum at 250 °C for 12 h and cooled to
77 K to collect the D2 isotherm. The closed and open
symbols in the inset refer, respectively, to adsorption and
desorption isotherms. The isotherm for D2 on the same
sample confirms the high storage observed for H2 at low
pressure. These isotherms also reveal the effect of the
quantum mechanical zero point motion on the storage.
That is, since the observed wt% (D2) is more than twice
the wt% (H2), the D2 wt% storage is significantly larger
than the expected (×2) enhancement stemming from the
larger isotopic mass of D2. The zero point motion en-
hancement in the surface binding energy is approxi-
mately 1/2 the mean kinetic energy of vibration (〈KE〉) of
the molecule in the pore,29 and the 〈KE〉 depends on the
nature of the confinement of the molecule in the pore
(e.g., interstitial channel in the bundle, internal pore of
an individual nanotube). In principle, the size of the iso-
tope effect can give indirect information on pore volume
and shape.29
channels. Thus, only the internal surface of the nanotube
contributes significantly (in this bundle structure) to the
theoretical SSA. To the best of our knowledge, no one
has yet reported values exceeding approximately 400 m2/g
for SWNT materials. Furthermore, it is important to note
that our values for the SSA do not correlate with the wt%
H2 storage. In another study on H2 storage in high sur-
face area activated carbons to be reported later, we also
found that the wt% stored did not correlate with the
measured SSA.26 These low SSA values for the present
SWNT sample suggest that many of the internal pores in
a real SWNT material have gateways smaller than the
kinetic diameter of N2, approximately 0.36 nm.27 As
pointed out previously,28 these restricted gateways may
be associated with the presence of carboxylic acid and
other functional groups attached to the carbon atoms lo-
cated at the edge of holes in the tube walls or at the rim
of open tube ends. These latter gateways to the internal
pore of the SWNT are, in many cases, apparently small
enough to block entry of N2, but not H2.
In Fig. 3, we show low-temperature isotherms for five
samples labeled A, B, C, D, and E (a description of the
processing details for these samples is given in Table I).
The most remarkable feature of the H2 storage isotherms
The isosteric heat of adsorption (qst) is a measure of
the binding energy of the molecule to the surface. The
isosteric heat of adsorption is given by qst ס R d(ln
P)/d(1/T), where R is the gas constant and P is the pres-
sure.27 We can estimate qst using data from two iso-
therms collected at T ס 77 and 87 K (see Fig. 3). For our
highest storage sample (E), we find qst ס 125 5 meV.
The isosteric heat bears a simple relation to the binding
energy. For example, a low-density gas on a planar sur-
face obeys the relation qst ס 〈Ez〉 + (3/2)kT, where 〈Ez〉
is the mean binding energy related to displacement of the
molecule normal to the surface. At T ס 77 K, kT ∼
6 meV, so qst ӷ kT, and we expect that the isosteric heat
and the binding energy are nearly equal. The magnitude
of qst substantially exceeds that expected for adsorption
FIG. 3. Liquid-nitrogen (T ס 77 K) isotherms showing progressively
better wt% H2 storage with nanotube processing. From top to bottom,
the isotherms were taken on as-received material (a) and after proc-
essing steps (b) through (d) were carried out (see caption to Fig. 1
and Table I). Samples (A) and (C) were degassed in vacuum in situ
in the TGA at approximately 300 °C for 12 h; for (B) and (D), the
sample was heated to 1000 °C. (E) Isotherms at T ס 87 K for
the same sample processed through step (d) (see caption to Fig. 1).
Inset: Comparison of the T ס 77 K isotherms for H2 and D2 in the
same sample, but a different sample used to collect the H2 isotherms
in Fig. 3. The wt% storage for D2 is larger than twice that for H2;
this additional storage is identified with a “zero-point” motion effect
(see text).
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2213
http://journals.cambridge.org
Downloaded: 11 Mar 2015
IP address: 169.230.243.252

B.K. Pradhan et al.: Large cryogenic storage of hydrogen in carbon nanotubes at low pressures
of H2 to a planar sheet of graphite (approximately 50 meV).
This observation motivates a consideration of geometry-
specific mechanisms for enhancing adsorption.
C. Nanotube damage studies
The end result of our aggressive HNO3 treatment is a
heavily damaged tube covered with holes of various
sizes. If all the C atoms were present in the tube wall
(i.e., no holes), a perfect wall structure built from hex-
agonal rings is obtained. A defective wall structure,
with missing C atoms, is far from equilibrium; the system
can gain stability by partial healing the holes during the
posttreatment anneal. The annealed tubes are likely to
have irregular shapes and roughened surfaces (as we
show below). Atom transport during the annealing of
the tube wall must, of geometrical necessity, produce
pentagonal and heptagonal carbon rings. Pentagonal
rings are particularly favored, as they most efficiently
cap bonds on edge atoms bordering the holes. Such ir-
regularities and surface roughness might open the small
interstitial channels of pristine bundles into wider inter-
stitial galleries. The disorder-induced dilation is particu-
larly important since the channels in a pristine tube
bundle are near the minimum size that still allow the
entrance of hydrogen.30
A roughened graphitic surface might also exhibit a
more subtle enhancement in adsorption arising from the
pervasive distortions in local bonding. We have at-
tempted to model adsorption on disordered graphene-like
structures by generating candidate structures and study-
ing the subtle energetics differences in hydrogen-binding
energy to their surfaces. Figure 4(a) shows the result of a
tight-binding molecular dynamics31 simulation of a de-
fective (8,8) nanotube. Ten percent of the atoms was first
removed in random patches from a pristine (8,8) struc-
ture. The system was then annealed at 2500 K (artifi-
cially high) and then cooled to the local equilibrium
structure over 1.5 ps. Since molecular dynamics can ad-
dress the system evolution only over short time scales,
the resulting structure in Fig. 4(a) should be viewed as
illustrative rather than definitive. A cross-sectional view
shows strong deviations from circularity. Figure 4(c) de-
picts a bundle of seven such tubes to demonstrate how
the roughening of the tube surfaces could enlarge the
interstitial channels. These seven tubes were relaxed as
individual rigid objects into a bundle geometry via inter-
tube van der Waals interactions. We believe that the re-
laxation of the internal coordinates within a tube would
have increased the interfacial contact area only slightly.
We next performed first-principles total-energy calcu-
lations within the local density approximation (LDA) for
hydrogen molecules weakly bound to the disordered tube
at various local structures present in the disordered
tube. We examined several model systems: (i) a 20-atom
section extracted from a disordered tube such as Fig. 4(a)
FIG. 4. (a) Computed nanotube structure with rough walls obtained
from simulation of a rapid high-temperature anneal of a damaged
tube, as described in the text. The starting structure was an (8,8) tube.
Darker carbon atoms exhibit larger deviations from planarity, showing
that the distortions propagate outward from pentagonal rings. A patch
extracted from such a nanotube is shown in (b), with the relaxed
position of a hydrogen molecule above the most distorted carbon atom.
The seven-tube bundle shown in cross section in (c) is built from the
same tube, but each replica has a random orientation and was allowed
to relax to an intertube separation dictated by a standard intertube van
der Waals interaction.
2214
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 11 Mar 2015
IP address: 169.230.243.252

B.K. Pradhan et al.: Large cryogenic storage of hydrogen in carbon nanotubes at low pressures
and capped at the edges by hydrogen atoms; (ii) a
few atmospheres. Previous reports in bundled SWNT of
8 wt% H2 storage at 77 K were found at considerable
higher pressure (approximately 100 atm). The remark-
able downshift we observed in the working H2 overpres-
sure (to approximately 2 atm) to achieve approximately
6 wt% H2 storage is not understood. Results of molecular
dynamics studies are presented that suggest there should
be an enhancement in the H2 binding energy associated
with a roughened nanotubes surface. The theoretical
binding energies for model rough surfaces are signifi-
cantly larger than for a planar graphene sheet but still much
smaller than the experimental value for the isosteric heat.
16-carbon atom planar molecule comprising four fused
hexagonal rings (denoted “6666”); (iii) a series of dis-
torted 14-carbon atom aromatic molecules composed
of two hexagons and two pentagons fused (“5665”). The
6666 molecule was fully relaxed; the 5665 series covered
a range of deviations from planarity to investigate the
heterogeneity of a roughened tube. In each case, we
placed a hydrogen molecule above the center of the car-
bon structure and allowed the adsorbate to relax. (We
also tested several different starting positions for the hy-
drogen molecule to more thoroughly explore possible
resting positions). The 6666 system, which models a flat
undistorted graphitic sheet, showed a H2 binding energy
0.072 eV, consistent with previous results for adsorption
on graphitic sheets.32 Both the 5665 series and the ex-
tracted patch from the disordered nanotube showed sub-
stantially increased adsorption energies: from 0.087 to
0.092 eV for the 5665 series; 0.095 eV for the extracted
nanotube patch.
To obtain an accurate measure of the actual enhance-
ment, these preliminary values were corrected for the
well-known LDA over binding of weakly bound molecu-
lar systems. To this end, we have computed accurate
first-principles diffusion Monte Carlo (DMC) energet-
ics33 for a system of three fused pentagons, obtaining
0.09 and 0.07(1) eV in LDA and DMC, respectively. The
characteristic LDA overbinding in this case is moderate
(approximately 20 meV), which lends confidence to the
overall picture. With application of an approximately
20-meV overbinding correction, the adjusted binding en-
ergies range from approximately 52 meV (in 6666) to
approximately 75 meV (in the 20-atom patch from the
disordered tube). The computed enhancement in binding
energy due to surface roughening is found to be quite
substantial, i.e., 40–50%. The local roughening seems to
affect the hydrogen binding through a complex interplay
of aromaticity, bond distortion, and local coordination.
These modeling results suggest that the local hydrogen
physisorption binding energy can be substantially en-
hanced through distortions arising from nanometer-scale
structural constraints. Note that the physisorption energy
increases even though the carbon coordination of a hy-
drogen molecule decreases due to the local convexity of
the carbon sheet.
ACKNOWLEDGMENTS
P.C.E. acknowledges financial support from the Na-
tional Science Foundation, Pennsylvania State Univer-
sity, Materials Research Science and Engineering Center,
Office of Naval Research under Grant No. N00014-99-
1-0252. J.C.G. acknowledges the support of the United
States Department of Energy by the University of Cali-
fornia Lawrence Livermore National Laboratory under
Contract No. W-7405-Eng-48. V.H.C. and D.S. acknowl-
edge the Army Research Office under Grant No.
DAAD19-99-1-0167.
REFERENCES
1. G.D. Berry and S.M. Aceves, Energy Fuels 12, 49 (1998).
2. S. Hynek, W. Fuller, and J. Bentley, Int. J. Hydrogen Energy 22,
601 (1997).
3. F. Rodriguez-Reinoso and A. Linares-Solano, Chemistry and
Physics of Carbon, edited by P.A. Thrower (Marcel Dekker, Inc.,
New York, 1989), Vol. 22, p. 1.
4. R.K. Agarwal, J.S. Noh, J.A. Schwarz, and P. Davini, Carbon 25,
219 (1987).
5. A.C. Dillon, K.M. Jones, T.A. Bekkedahl, C.H. Kiang,
D.S. Bethune, and M.J. Heben, Nature 386, 377–379 (1997).
6. M.J. Heben, A.C. Dillon, T. Gennett, J.L. Alleman, P.A. Parilla,
K.M. Jones, and G.L. Hornyak, in Nanotube and Related
Materials, edited by A.M. Rao (Mater. Res. Soc. Symp. Proc.,
633), p. 17.
7. C. Liu, Y.Y. Fan, M. Liu, H.T. Cong, H.M. Cheng, and
M.S. Dresselhaus, Science 286, 1127–1129 (1999).
8. P. Chen, X. Wu, J. Lin, and K.L. Tan, Science 285, 91–93 (1999).
9. C. Nutzenadel, H. Zuttel, D. Chartouni, and L. Schlaphach, Elec-
trochem. Solid State Lett. 2, 30 (1999).
10. N. Rajalakshmi, K.S. Dhathathreyan, A. Govindaraj, and
B.C. Satishkumar, Electrochim. Acta 45, 4511 (2000).
11. S.M. Lee, K.S. Park, Y.C. Choi, Y.S. Park, J.M. Bok, D.J. Bae,
K.S. Nahm, Y.G. Choi, S.C. Yu, N-g. Kim, T. Frauenheim, and
Y.H. Lee, Synth. Met. 113, 209 (2000).
IV. CONCLUSION
In conclusion, we have shown experimentally that the
wt% hydrogen storage in bundles of nanotubes at 77 K
falls in the range from approximately 1–6 wt% at ap-
proximately 2 atm, depending on the postsynthesis proc-
essing of SWNT materials. Aggressive HNO3 refluxing,
followed by a high T ∼ 1000 °C vacuum anneal (approxi-
mately 20 h) produces the highest storage >6 wt% at a
12. A. Chambers, C. Park, R.T.K. Baker, and N.M. Rodriguez,
J. Phys. Chem. B 102, 4253 (1998).
13. Y. Ye, C.C. Ahn, C. Witham, B. Fultz, J. Liu, A.G. Rinzler,
D. Colbert, K.A. Smith, and R.E. Smalley, Appl. Phys. Lett. 74,
2307 (1999).
14. R.T. Yang, Carbon 38, 623 (2000).
15. F.E. Pinkerton, B.G. Wicke, C.H. Olk, G.G. Tibbetts, G.P. Meisner,
M.S. Meyer, and J.F. Herbst, J. Phys. Chem. B 104, 9460 (2000).
J. Mater. Res., Vol. 17, No. 9, Sep 2002
2215
http://journals.cambridge.org
Downloaded: 11 Mar 2015
IP address: 169.230.243.252

B.K. Pradhan et al.: Large cryogenic storage of hydrogen in carbon nanotubes at low pressures
16. C.C. Ahn, Y. Ye, B.V. Ratnakumar, C. Witham, R.C. Bowman, Jr.,
25. Hiden User Manual.
and B. Fultz, Appl. Phys. Lett. 73, 3378 (1998).
17. B.K. Gupta and O.N. Srivastava, Int. J. Hydrogen Energy 25, 825
(2000).
26. B.K. Pradhan and P.C. Eklund (unpublished result).
27. S.J. Gregg and K.S.W. Sing, Adsorption, Surface Area and Poros-
ity (Academic Press, San Diego, CA, 1982).
18. M.S. Dresselhaus, K.A. Williams, and P.C. Eklund, MRS Bull. 45
(1999).
28. D.B. Mawhinney, V. Naumenko, A. Kuznetsova, J.T. Yates, Jr.,
J. Liu, and R.E. Smalley, Chem. Phys. Lett., 324, 213 (2000).
29. M.K. Kostov, M.W. Cole, J.C. Lewis, D. Phong, and J.K. Johnson,
Chem. Phys. Lett. 332, 26 (2000).
30. M.M. Calbi, F. Toigo, and M.W. Cole, Phys. Rev. Lett. 86, 5062
(2001).
31. D. Frankel and B. Smit, Understanding Molecular Simulation:
From Algorithms to Applications (Academic Press, San Diego,
CA, 1996); D. Porezag, Th. Freuenheim, and Th. Kohler, Phys.
Rev. B 51, 12947 (1995).
19. A.C. Dillon and M.J. Heben, Appl. Phys. A 72, 133 (2001);
V. Meregalli and M. Pariivello, Appl. Phys. A 72, 143 (2001);
M. Hirscher, M. Becher, M. Haluska, U. Dettlaff-Weglikowska,
A. Quintel, G.S. Duesberg, Y-M. Choi, P. Downes, M. Hulman,
S. Roth, I. Stepanek, and P. Bernier, Appl. Phys. A 72, 129
(2001); L. Schlapbach and A. Zuttl, Nature 414, 353 (2001).
20. CarboLex Inc.; web site: www.carbolex.com.
21. M.S. Dresselhaus and P.C. Eklund, Adv. Phys. 49, 705 (2000).
22. M.S. Dresselhaus, G. Dresselhaus, and P. C. Eklund, Science of
Fullerenes and Carbon Nanotubes (Academic Press, San Diego,
CA, 1996).
23. R. Saito, G. Dresselhaus, M.S. Dresselhaus, Physical Properties
of Carbon Nanotubes (Imperial College Press, Singapore, 1998).
24. C. Journet, W.K. Maser, P. Bernier, A. Loiseau, M. Lamy
de la Chapelle, S. Lefrant, P. Deniard, R. Lee, and J.E. Fischer,
Nature 388, 756 (1997).
32. J.S. Arellano, L.M. Molina, A. Rubio, and J.A. Alonso, J. Chem.
Phys. 112, 8114 (2000).
33. A.J. Williamson, R.Q. Hood, and J.C. Grossman, Phys. Rev. Lett.
87, 246406 (2001).
2216
J. Mater. Res., Vol. 17, No. 9, Sep 2002
http://journals.cambridge.org
Downloaded: 11 Mar 2015
IP address: 169.230.243.252