Enhancing low pressure hydrogen storage in sodium alanates

Article abstract of DOI:10.1016/S0925-8388(01)01940-5

We investigated the dehydrogenation of NaAlH₄ and the reversible low-pressure rehydrogenation from NaH to Na₃AlH₆. Highly purified NaAlH₄ requires relatively high temperatures to decompose to NaH and long durations to rehydride to the Na₃AlH₆ phase in hydrogen gas. However, any degradation of the purity of this material, whether through ball milling with diamond powder or ball milling with diamond plus Al powders, mixing the purified material with Pt powder, or doping with a Ti organometallic compound, lowers the decomposition temperature and facilitates rehydriding the product NaH+Al to Na₃AlH₆. Diamond ball milling of NaAlH₄ seems to be the best of these procedures; it substantially decreases the decomposition temperatures, with significant dehydrogenation starting at 180°C rather than 250°C for the purified material, and with formation of NaH substantially complete at 235°C rather than 290°C. Rather surprisingly, it also facilitates rehydrogenation from NaH+Al to Na₃AlH₆. Similarly, NaAlH₄ doped with Ti according to the recipe of Bogdanovic? lowers the decomposition temperatures and improves the hydrogenation kinetics for the low pressure transition from NaH+Al to Na₃AlH₆. Pressure-composition isotherms show that the rehydrogenation of the resulting NaH+Al decomposition phases into the Na₃AlH₆ intermediate phase at pressures below 3.6 MPa is similar for the diamond ball milled and Ti-doped material. Diamond ball milling NaAlH₄ with excess Al did not improve the rehydrogenation kinetics.

Full text of DOI:10.1016/S0925-8388(01)01940-5

Journal of Alloys and Compounds 337 (2002) 254–263
L
www.elsevier.com/locate/jallcom
Enhancing low pressure hydrogen storage in sodium alanates
*
G.P. Meisner , G.G. Tibbetts, F.E. Pinkerton, C.H. Olk, M.P. Balogh
GM Research and Development Center, Materials and Processes Laboratory, MC 480 106 224, Warren, MI 48090-9055, USA
Received 12 October 2001; accepted 29 October 2001
Abstract
We investigated the dehydrogenation of NaAlH4 and the reversible low-pressure rehydrogenation from NaH to Na3AlH6. Highly
purified NaAlH4 requires relatively high temperatures to decompose to NaH and long durations to rehydride to the Na3AlH6 phase in
hydrogen gas. However, any degradation of the purity of this material, whether through ball milling with diamond powder or ball milling
with diamond plus Al powders, mixing the purified material with Pt powder, or doping with a Ti organometallic compound, lowers the
decomposition temperature and facilitates rehydriding the product NaH1Al to Na3AlH6. Diamond ball milling of NaAlH4 seems to be
the best of these procedures; it substantially decreases the decomposition temperatures, with significant dehydrogenation starting at 180 8C
rather than 250 8C for the purified material, and with formation of NaH substantially complete at 235 8C rather than 290 8C. Rather
surprisingly, it also facilitates rehydrogenation from NaH1Al to Na3AlH6. Similarly, NaAlH4 doped with Ti according to the recipe of
´
Bogdanovic lowers the decomposition temperatures and improves the hydrogenation kinetics for the low pressure transition from
NaH1Al to Na3AlH6. Pressure–composition isotherms show that the rehydrogenation of the resulting NaH1Al decomposition phases
into the Na3AlH6 intermediate phase at pressures below 3.6 MPa is similar for the diamond ball milled and Ti-doped material. Diamond
ball milling NaAlH4 with excess Al did not improve the rehydrogenation kinetics.
2002 Elsevier Science B.V. All rights reserved.
Keywords: Hydrogen storage materials; Gas–solid reaction; X-Ray diffraction
1. Introduction
3NaAlH4↔Na3AlH6 1 2Al 1 3H2 3.70 wt.% H2
(1)
(2)
The major automobile companies are now developing a
fuel cell for automotive use that, together with a hydro-
carbon reformer, may efficiently power methanol or
gasoline fueled vehicles [1] in the intermediate term
(perhaps through 2010). In the long term (beyond 2010), a
‘hydrogen economy’ infrastructure might be built to
provide zero emission fuel cell vehicles powered with
hydrogen. In either scenario, it will be necessary to store
some hydrogen on board the vehicle; energy efficiency will
require that it store over 6.5 wt.% hydrogen [2] which can
be released at temperatures below 150 8C.
This need for hydrogen storage media has accelerated
research in this field. We have recently come to negative
conclusions about the suitability of carbon-based nanofib-
ers and nanotubes [3] in agreement with some recent work
[4] and despite many recent claims [5].
Na3AlH6 1 2Al 1 3H2↔3NaH 1 3Al 1 9/2H2
1.85 wt.% H2
providing a total of 5.55 wt.% hydrogen at the NaH
endpoint, it was not known how to make these decomposi-
tions reversible. However, a 1997 paper by Bogdanovic
and Schwickardi [6] awakened interest by showing how
reversibility could be catalyzed with transition metals,
particularly Ti.
Recently, Zidan et al. [7] reported that doping NaAlH4
with both Ti and Zr lowered the dehydriding temperatures
of both stages of decomposition, offering measurable
dehydriding below 100 8C. Zaluska et al. [8] succeeded in
accelerating the dehydrogenation kinetics by ball milling
alanates without transition metal catalysts and by ball
milling with the addition of carbon. Nevertheless, the
decomposition temperatures required are still high for
automotive applications. Despite providing a somewhat
lowered temperature for the onset of dehydrogenation,
temperatures approaching 200 8C were still required for
complete hydrogen removal described by Eqs. (1) and (2).
Furthermore, the extensive processing required to properly
´
Although NaAlH4 has been known for many years for
its two-step thermal decomposition,
*
Corresponding author.
E-mail address: gregory.p.meisner@gm.com (G.P. Meisner).
0925-8388/02/$ – see front matter
PII: S0925-8388(01)01940-5
2002 Elsevier Science B.V. All rights reserved.

G.P. Meisner et al. / Journal of Alloys and Compounds 337 (2002) 254–263
255
distribute dopants throughout the material could be a
significant disadvantage. If grain size reduction can pro-
vide comparable results, grinding methods will be very
attractive. A recent article by Jensen and Gross [9]
compares enhanced dehydrogenation results obtained by
catalysis and doping.
In this paper we explore several of these avenues. We
compare the accelerated rates of dehydrogenation observed
due to grain refinement to those observed for Ti-doping.
We also explore the notion of improving the reversibility
of NaH hydrogenation by adding surplus Al. By carefully
comparing the rates of sorption and desorption of samples
treated by these processes with the pure alanate, we can
quantify the sorption rate improvements obtained.
filtered and dried under vacuum overnight at room tem-
perature [6]. After drying, the NaAlH4 was obtained as a
fine white powder.
The Ti-doped alanate sample was prepared using
Ti[O(CH2)3CH3]4. The NaAlH4 was suspended in 200 ml
of toluene and stirred at room temperature for 5 h. Then 2
mol% of Ti[O(CH2)3CH3]4 was added and the mixture
stirred at 35 8C overnight. Finally, the toluene was evapo-
rated and the residual solid dried overnight under vacuum.
A sample containing 10 wt.% Pt was prepared by
superficial doping: black ‘amorphous’ Pt powder (0.15–
0.45 mm particle size, purchased from Alfa Aesar) was
added to the white purified NaAlH4 powder in the Pt
measurement pan of the thermogravimetric analyser and
stirred with a metal implement until the mixture was a
homogeneous gray color.
2. Experimental
X-Ray powder diffraction data were collected using a
Bruker AXS General Area Detector Diffractometer
equipped with Cu Ka radiation. We loaded samples into
1.5 mm glass capillary tubes and sealed them in an inert
atmosphere glove box prior to the diffraction measure-
ments. The observed Debye rings were integrated to give
powder diffraction spectra. In some cases, however, the
Debye rings were incomplete, consisting of a series of
spots because of the relatively large crystallite size com-
pared to the capillary diameter.
Because of the reactivity of the alanates, every attempt
was made to handle the various samples in inert atmos-
phere environments in all procedures and measurements,
the notable exception being the scanning electron micro-
scopy (SEM) work discussed below.
2.1. Sample preparation and characterization
The NaAlH4, purchased from Lancaster Synthesis,
nominally 90% pure, was used as-received for preparing
the ball milled samples. Ball milling was carried out in a
Spex 8000 steel mixer mill using two 1.27 cm diameter
and four 0.635 cm diameter steel balls of aggregate mass
of 21 g for each 2 g of sample. The first sample contained
20 wt.% of 6 mm diamond powder mixed with 80 wt.% of
NaAlH4 and was milled continuously for 1 h. In an attempt
to reduce kinetic limitations on the rehydriding of NaH, we
ball milled a second sample of NaAlH4 with 20 wt.% Al
powder (|99% pure Al, obtained from Aldrich) in addition
to 10 wt.% diamond powder. To ensure that heating and
partial desorption of the hydrogen was minimized, the
second sample was milled in 5-min intervals followed by
30 min cooling for a total milling time of |3 h. The bulk
of the sample remained cooler than 100 8C although impact
point temperatures were probably considerably higher.
The purified alanate samples used in this study were
prepared as follows. A 1.0 M solution of commercial
NaAlH4 in THF (tetrahydrofuran) was prepared using the
Shlenk method [10]. Typically, 10 g of NaAlH4 in 100 ml
of THF was stirred for 1 h and passed through a glass
filter. The filtrate was then concentrated in vacuum to a
volume of |30 ml where NaAlH4 began to crystallize and
separate from the solution. A 100-ml volume of toluene
was added to the THF solution and the mixture was stirred
for 3 h under an argon atmosphere, causing NaAlH4 to
fully separate from the solution as a fine precipitate. The
suspension was stirred for two more hours and filtered. The
residue was washed with toluene three times, and then
Examination of these materials in our Hitachi S-4000
field emission SEM system was not straightforward. We
utilized two-sided carbon tape to adhere the powders to the
SEM discs. Our first attempt to view these materials failed
when samples began to deliquesce and bubble as they were
being transported in air from the sample storage glove box
to the SEM system. As one might expect, the finely ground
materials were more reactive than the large crystallites. We
minimized the problem by attaching the discs to SEM
specimen mounts in the glove box and transporting these
assemblies to the SEM room in screw topped jars filled
with Ar. Nevertheless, all samples were exposed to air for
several seconds as the mounts were removed from the jars,
mounted in the SEM system, and pumped down.
2.2. Hydrogen sorption measurements
We measured the thermal desorption properties of
sodium alanates using two types of experimental equip-
ment: (1) a computer-controlled commercial Gas Reaction
Controller (GRC) manufactured by the Advanced Materi-
als Corporation of Pittsburgh, PA, USA, described previ-
ously [3], and (2) a Perkin-Elmer System 7 thermogravi-
metric analysis (TGA) system located inside an Ar glove
box.
For the GRC measurements, samples weighing |1 g
were heated in the evacuated sample chamber to |300 8C
at either 1 or 5 8C/min, and the amount of desorbed
hydrogen gas was determined from gas pressure measure-
ments. Because of the relatively large temperature gradient

256
G.P. Meisner et al. / Journal of Alloys and Compounds 337 (2002) 254–263
present in the GRC (e.g. |1.25 8C/mm measured vertically
in the sample holder at 200 8C when heating at 1 8C/min),
the GRC thermal desorption curves are considerably
broader and shifted in temperature compared to the TGA
data. We believe that the TGA results are the more accurate
determination of the thermal desorption profile of these
materials.
For the TGA measurements, samples weighing from 5
to 14 mg were placed onto the Pt microbalance pan, the
system was flushed with Ar, and the sample was then
heated at 1 8C/min from 30 8C to 300 8C in a 45 sccm flow
of Ar. At temperatures below 160 8C, the samples showed
differing behavior: all samples except the ball milled
material lost small amounts of mass, which we attribute to
solvent loss. The ball milled sample actually gained
weight, perhaps because the newly exposed surfaces
scavenged oxygen and/or water vapor impurities from the
TGA gas flow with greater efficiency. To minimize these
small errors we used the initial mass of each sample at
160 8C as the starting mass. The GRC measurement is
much less sensitive to these effects, and no background
correction for gas desorption was applied to any of the
GRC data.
Pressure–composition (P–C) isotherms were measured
(in situ after the initial heating to 300 8C in the GRC
described above) by evacuating the GRC sample chamber
and holding the sample at various fixed temperatures. P–C
isotherm measurements as a function of increasing pres-
sure consist of the apparatus (1) causing hydrogen gas to
fill a reservoir, (2) measuring the equilibrium temperature
and pressure of the reservoir to determine the number of
gas molecules present, (3) admitting aliquots of hydrogen
gas from the reservoir to the sample chamber, (4) measur-
ing new equilibrium pressures and temperatures of the
reservoir and sample chamber to determine the number of
molecules of hydrogen remaining in the gas phase, (5)
interpreting any decrease in this number of gas molecules
as hydrogen sorbed by the sample, (6) gradually increasing
the hydrogen pressure to the maximum allowed, about 4
MPa, while monitoring the sorbed hydrogen. Similarly, the
amount of hydrogen released from the sample is then
determined by pumping out the reservoir and gradually
removing fixed amounts of hydrogen gas from the sample
chamber. After each aliquot of hydrogen gas is transferred
from the sample chamber to the reservoir, measurements of
equilibrium pressure and temperature determine the num-
ber of additional molecules of hydrogen present in the gas
phase, and any extra gas molecules correspond to hydrogen
desorbed from the sample.
Thus for a given sample and value of t, the total data
acquisition time ot for pressurization and depressurization
can vary by nearly an order of magnitude. In our experi-
ments t was typically 100 s, but t values ranging from 30
to 1500 s were used as well. As we show in the results
below, the kinetics of the alanates can be so slow that even
in experiments with t51500 s (and ot.180 h) equilib-
rium may not be reached.
3. Results
Initial desorption curves were measured by GRC and
TGA, and both gave qualitatively the same results. Fig. 1
shows the weight loss as a function of temperature for the
Lancaster Synthesis as-received, diamond ball milled,
purified, and purified doped with 10 wt.% Pt, where the
data for the diamond ball milled and Pt-doped samples are
corrected for the diamond and Pt content, respectively.
Both as-received and ball milled materials clearly show
desorption curves characteristic of a two-step desorption
process as described by Eqs. (1) and (2) above. The first
desorption for the as-received material is completed below
210 8C and the second occurs between about 250 and
270 8C. After ball milling, the first desorption becomes
very sharp near 180 8C and the second is substantially
complete at 235 8C. In contrast, the data for the purified
and Pt-doped purified materials do not show a clear two-
step desorption behavior. The purified sample shows only
a little desorption before 210 8C, with nearly all of the
desorption occurring between 200 8C and 300 8C. The
One of the parameters the user can set for a P–C
isotherm measurement using the GRC is a time constant t
for the amount of time the apparatus waits between
pressure gauge readings. During hydrogen sorption, the
apparatus typically records a data point only after many t
intervals and advances to the next pressure increment only
when sufficient equilibrium conditions have been reached.
Fig. 1. TGA and GRC desorption curves for NaAlH4 samples heated at
1 8C/min: purified, purified and lightly mixed with Pt powder, as-received
Lancaster Synthesis, powder nominally 90% pure, and ball milled
Lancaster Synthesis, powder using 20 wt.% diamond powder for 1 h. The
data for the diamond milled and Pt-doped materials have been scaled to
correct for NaAlH4 phase dilution by the 20 wt.% diamond and 10% Pt,
respectively.

G.P. Meisner et al. / Journal of Alloys and Compounds 337 (2002) 254–263
257
Pt-doped purified sample shows a thermal desorption
profile that contains only subtle features suggestive of the
two-step desorption process. It is shifted by 20 8C to lower
temperatures compared to the purified (undoped) material,
although the decomposition is at higher temperatures than
observed for the as-received and ball milled samples.
The amounts of hydrogen desorption determined from
the TGA weight loss measurements and from the GRC
pressure measurements for the various materials studied
are listed in Table 1 along with the theoretical maximum
hydrogen desorption based on Eqs. (1) and (2). The data
for the as-received, diamond ball milled for two heating
rates (1 and 5 8C/min), and diamond1Al ball milled
samples all show two clearly identifiable stages of de-
composition, although the total mass loss is much less than
the theoretical maximum. The total mass losses from the
purified and Pt-doped purified samples do agree well with
the theoretical maximum hydrogen loss in going from
NaAlH4 to NaH, but the losses are not clearly assignable
to two stages as in Eqs. (1) and (2).
X-Ray diffraction spectra, Fig. 2, and scanning electron
micrographs, Fig. 3, were obtained for the following
samples made with NaAlH4 purchased from Lancaster
Synthesis: (a) as-received; (b) ball milled with 20 wt.%
diamond, before heating; (c) ball milled with 20 wt.%
diamond, after heating rapidly to 170 8C, soaking for 2.75
h, and cooling to room temperature in hydrogen; and (d)
ball milled with 20 wt.% diamond, heated slowly to 300 8C
and cooled to room temperature in vacuum. In Fig. 2a, the
split diffraction peaks arise from incomplete Debye rings
(as discussed in the Experimental section above) whereas
the shape of the diffraction peaks for the three balled
milled samples, Fig. 2b–d, is more uniform because the
finer, ball milled particles in these samples diffract more
continuous and complete Debye rings.
Fig. 2. X-Ray diffraction spectra for samples prepared from NaAlH4
(Lancaster Synthesis, nominally 90% pure): (a) as-received; (b) after ball
milling with 20 wt.% diamond, before heating; (c) after ball milling with
20 wt.% diamond, heated rapidly to 170 8C, soaked for 2.75 h, and cooled
to room temperature in hydrogen; and (d) after ball milling with 20 wt.%
diamond, heated slowly to 300 8C and cooled to room temperature in
vacuum.
average grain size is reduced, but it does not cause any
further decomposition judging by the relative diffraction
intensity associated with each phase. In other words,
except for the addition of the diamond, the alanate phase
content of the as-received and diamond ball milled materi-
als are the same. Sustained heating at 170 8C, however,
decomposes the NaAlH4 phase, giving predominantly
Na3AlH6 and Al, with a trace amount of NaAlH4 but no
NaH, Fig. 2c. Finally, heating the sample to 300 8C results
in nearly complete conversion to NaH and Al, Fig. 2d,
with only a trace amount of Na3AlH6. Even when the
amount of hydrogen desorbed by the ball-milled sample is
corrected for the diamond addition, it is considerably
below the 5.55 wt.% expected from Eqs. (1) and (2), again
consistent with some prior decomposition of the as-re-
ceived material as indicated by the X-ray diffraction data
above and the desorption data listed in Table 1. It is also
possible that the ball milling step causes some hydrogen
release and corresponding decomposition products that are
not crystallized well enough to be seen in the diffraction
The X-ray diffraction spectra for both the as-received
sample, Fig. 2a, and the diamond ball milled, unheated
sample, Fig. 2b, show predominantly the NaAlH4 phase
with very small but observable amounts of impurity phases
Na3AlH6 and Al. We presume this represents some prior
decomposition of NaAlH4 in the as-received state. Ball
milling causes broadening of the diffraction peaks as the
Table 1
Hydrogen desorption results from TGA and GRC measurements
Material
TGA
GRC
Dm1
Dm2
Dmtotal
Dm1 /Dm2
Dm1
Dm2
Dmtotal
Dm1 /Dm2
(wt.%)
(wt.%)
(wt.%)
(wt.%)
(wt.%)
(wt.%)
Theoretical maximum
3.70
1.85
5.55
2.0
3.70
1.85
5.55
2.0
As-received
3.2
2.0
1.6
1.5
4.8
3.5
2.0
1.3
Diamond milled, 1 8C/min
Diamond milled, 5 8C/min
Diamond1Al milled
Purified
2.94
2.86
2.2
1.59
1.71
1.5
4.53
4.57
3.7
1.85
1.67
1.5
5.5
5.5
5.1
Purified1Pt

258
G.P. Meisner et al. / Journal of Alloys and Compounds 337 (2002) 254–263
Fig. 3. Scanning electron micrographs of samples prepared from NaAlH4 (Lancaster Synthesis, nominally 90% pure): (a) as-received; (b) after ball milling
with 20 wt.% diamond, before heating; (c) after ball milling with 20 wt.% diamond, heated rapidly to 170 8C, soaked for 2.75 h, and cooled to room
temperature in hydrogen; and (d) after ball milling with 20 wt.% diamond, heated slowly to 300 8C and cooled to room temperature in vacuum.
data. A recent series of in-situ X-ray diffraction measure-
ments by Gross et al. [11] detailed the phase changes
expected on heating NaAlH4.
remaining from the decomposition of NaAlH4. We ob-
served that the rehydriding of NaH1Al to Na3AlH6 could
be accomplished with very little capacity loss for at least
15 rehydriding cycles.
The scanning electron micrographs show that ball
milling with diamond particles pulverizes the 100 mm
crystallites of the as-received material, Fig. 3a, to irregular
particles whose diameter is about 2 mm, Fig. 3b, although
it may be that deliquescence has cemented these particles
together from the 0.3 mm granules evident in Fig. 3b–d.
The particles have numerous 0.1 mm pores that we assume
arise from hydrogen transport through the crystallites. This
morphology does not seem to change after heating to
170 8C, Fig. 3c, or even beyond the presumed melting
transition [11] to 300 8C, Fig. 3d.
We extensively studied the sorption and desorption of
hydrogen in these materials in the low-pressure range of
0–4 MPa where the reverse reaction of Eq. (2) is known to
lead to hydrogenation of NaH. As Fig. 2c shows, this
intermediate compound is Na3AlH6. Theoretically, this
low pressure sorption can incorporate a maximum of 1.85
wt.% hydrogen, assuming that no NaAlH4 is produced and
that one begins the sorption with all of the products
In Fig. 4, we show P–C isotherms measured at 201 8C
for the diamond ball milled sample after having been
heated to 300 8C. The three curves were collected with
t530, 100, and 600 s, corresponding to ot53.4, 13.7,
and 80.1 h, respectively. In the 3.4 h experiment, the slow
kinetics of the rehydrogenation of NaH is evident both in
the high plateau pressure and low maximum capacity
achieved. With ot513.7 h (t5100 s), however, the
experiment is evidently long enough to achieve maximum
capacity because the sample reaches the same peak absorp-
tion as in the ot580.1 h experiment. Kinetic effects are
evident, however, because the plateau pressures required
for absorption and desorption are higher and lower,
respectively, for the 13.7 h compared to the 80.1 h
experiment.
Using t5100 s, the series of P–C isotherms shown in
Fig. 5 was measured for the diamond ball milled material.
These data were obtained using two different specimens

G.P. Meisner et al. / Journal of Alloys and Compounds 337 (2002) 254–263
259
Fig. 4. Pressure–composition isotherms at 201 8C with three different
values of t (ot) for NaAlH4 ball milled with 20 wt.% diamond powder
for 1 h and then heated to 300 8C in the GRC to fully decompose the
NaAlH4. The hydrogen content is weight percentage relative to the
alanate content, i.e. they have been corrected for 20 wt.% diamond.
Fig. 6. Pressure–composition isotherms with t5100
NaAlH4. The maximum ot was 7.7 h for the P–C isotherm obtained at
206 8C.
s
for purified
shows a much smaller capacity for re-hydrogenating
NaH1Al into Na3AlH6 (Fig. 6) compared to the diamond
ball milled material (Fig. 5). This is a kinetic effect,
because even with a 64 h data acquisition time shown in
Fig. 7 the capacity is only just approaching |1.1 wt.%,
very much less than the 1.85 wt.% expected according to
Eq. (2), and an experiment with t51500 s, giving a total
data acquisition time ot.180 h, only gives a capacity of
|1.2 wt.%, still less than the results for diamond ball
milled material.
The purified NaAlH4 is comprised of |20 mm crys-
tallites, Fig. 8a, and after all of our cycling experiments
and numerous heatings to 300 8C, the crystallites roughen
and show numerous pores that we assume arise from
hydrogen transport, Fig. 8b. Nevertheless the gross dimen-
from the same diamond ball milled batch; both specimens
were heated to 300 8C to fully decompose the NaAlH4
phase and then held at constant temperatures during the
P–C isotherm measurements. The data in Figs. 4 and 5
represent the reversible hydrogenation characteristics of
the Na3AlH6 phase. At the lowest temperature of 136 8C,
hydrogen is not able to diffuse into the lattice and bond. At
the intermediate temperature of 182 8C, hydrogen may
easily diffuse and bond within the solid so that maximum
sorption is obtained. At 227 8C, Na3AlH6 rapidly de-
composes, so sorption in the solid phase is small.
Fig. 6 shows P–C isotherms obtained under similar
conditions (i.e. t5100 s) for the purified material (not ball
milled) after heating to 300 8C. Here, the purified material
Fig. 5. Pressure–composition isotherms with t5100 s for diamond ball
milled NaAlH4. The maximum ot was 15.5 h for the P–C isotherm
obtained at 182 8C. These hydrogen content data represent the reversible
hydriding of NaH1Al to Na3AlH6 and are weight percentages relative to
the alanate content, i.e. they have been corrected for the 20 wt.%
diamond.
Fig. 7. Pressure–composition isotherms for purified NaAlH4 using t5
100, 500, and 1500 s, with the total data acquisition times shown in
parentheses. Here, the kinetics are much slower than for the diamond ball
milled material shown in Figs. 4 and 5.

260
G.P. Meisner et al. / Journal of Alloys and Compounds 337 (2002) 254–263
shows a set of P–C isotherms for the Ti-doped sample,
showing that at most 1.75 wt.% is sorbed at 3.6 MPa and
176 8C.
Finally, we conducted a thermal desorption profile
measurement of the diamond plus Al ball milled sample
using the GRC, and it is again very similar to that of the
diamond ball milled material. The mass losses corre-
sponding to the two-step desorption behavior were less
than those for the diamond ball milled sample and are
listed in Table 1. A series of P–C isotherm measurements
of the sample ball milled with diamond plus Al, however,
showed significantly reduced kinetics for hydrogen cycling
of the Na3AlH6 phase with the added Al even though the
crystallite size was essentially the same as for the sample
ball milled with diamond alone. Indeed, the kinetics of the
initial decomposition of the phases are similar in both ball
milled materials, but evidently additional macroscopic
grains of Al do not easily accelerate the rehydrogenation
reaction of Eq. (2).
We discuss these P–C isotherm results for the Na3AlH6
phase in more detail below for all the materials we have
studied.
4. Discussion
Fig. 1 shows alanate desorption stops at somewhat
higher temperatures, perhaps 30 8C, than those achieved by
Zidan et al. [7]. Our GRC measurements utilized apparatus
similar to that used in Ref. [7]: we ramped the furnace
slowly and linearly to 300 8C while measuring the pressure
increases due to hydrogen desorption. Careful studies have
convinced us, however, that there is an unacceptably large
temperature gradient across our sample holder. By com-
parison the TGA has been designed for good temperature
control of a much smaller sample and yields more accurate
measurements. Nevertheless, we observed substantially
similar thermal GRC desorption profiles for the diamond
ball milled, diamond plus Al ball milled, and Ti-doped
materials.
Differences in the total hydrogen desorption during
these initial warm-ups are consistent with different
amounts of prior decomposition of the NaAlH4 phase. In
other words, the various materials contain a mixture of
NaAlH4 and other phases. Such prior decomposition is
reflected in the ratio of Dm1 /Dm2, which equals 2.0 for
pure NaAlH4, i.e. no prior decomposition, and decreases
linearly with the amount of NaAlH4 converted to
Na3AlH6. From Table 1, one can see for example that the
slight decrease in total desorption, Dmtotal for the as-
received versus the diamond ball milled material (4.8
versus 4.53 wt.%) is consistent with the decrease in Dm1 /
Dm2 from 2.0 to 1.86. The significantly lower total
desorption of the ball milled material measured with the
TGA of 3.5 wt.% is evidence for further decomposition of
the NaAlH4 phase after the several months that elapsed
Fig. 8. Scanning electron micrographs of purified NaAlH4 (a) before
heating, (b) after heating to 300 8C and many hydriding cycles, and (c)
higher magnification of (b) showing that the gross dimensions of each
crystallite remain |20 mm with some crystal facets apparently undis-
turbed.
sions of each crystallite remain about 20 mm with some
crystal facets apparently undisturbed, Fig. 8c.
The thermal desorption profile of the Ti-doped sample
(not shown) was measured using the GRC and was very
similar to that of the diamond ball milled sample. Fig. 9

G.P. Meisner et al. / Journal of Alloys and Compounds 337 (2002) 254–263
261
Fig. 9. Pressure–composition isotherms with t5100 s for Ti-doped
NaAlH4. The maximum ot was 18.4 h for the P–C isotherm obtained at
176 8C. These data have not been corrected for the 2 wt.% Ti content.
between the GRC and the TGA measurements. Again, the
further decrease in Dm1 /Dm2 to 1.3 supports this conclu-
sion.
Fig. 11. A Van’t Hoff-type plot for various sodium alanates determined
from low pressure (,4 MPa) pressure–composition isotherms. Bog-
Fig. 10 compares the hydrogen storage capacities of the
Na3AlH6 phase at 3.6 MPa for the diamond ball milled,
diamond plus Al ball milled, purified, and Ti-doped
sodium alanates in the low pressure regime as a function of
temperature. Because we are constrained by our apparatus
to the pressure range below 4 MPa, this hydrogenation
corresponds to the reaction described by Eq. (2) and has a
maximum theoretical sorption of 1.85 wt.%. The data are
´
danovic refers to Ref. [12].
obtained by interpolation from the P–C isotherms in Figs.
5, 6, and 9, with the hydrogen storage capacities based on
the total weight of the initial sample of NaAlH4. The
hydrogen storage capacity results for the ball milled
materials are corrected in Fig. 10 to account for the fact
that the diamond ball milled and the diamond plus Al ball
milled materials contain only 80 and 70 wt.% NaAlH4,
respectively. (The hydrogen storage capacity for the Ti-
doped material was not corrected for the |2 wt.% Ti
content). It is encouraging that even though our X-ray
results show measurable initial decomposition of NaAlH4
after ball milling, and Table 1 underscores its loss of initial
capacity compared to stoichiometric material, ball milled
NaAlH4 may be recycled to near full Na3AlH6 capacity by
exposure to 3.6 MPa of hydrogen.
It obviously requires a much longer time to hydrogenate
the purified material with a hydrogen capacity reaching
only |1.2 wt.% for an experiment duration of over 180 h,
as we discussed above. Diamond ball milling quite effec-
tively removes this kinetic limitation with results remark-
ably similar to that of Ti-doping. Both yield a peak
sorption near 1.75 wt.% over the range 170–220 8C. The
Ti-doped material does seem to be preferred because of the
somewhat lower temperature where peak sorption is
reached, but perhaps it is most significant that both
samples reach a sorption near the theoretical maximum
with similar kinetics, i.e. they both reach equilibrium with
the experiment time constant t5100 s corresponding to a
maximum total data acquisition time ot,20 h.
Fig. 10. Hydrogen storage capacities at 3.6 MPa for various sodium
alanates determined from the P–C isotherm data. Each sample was first
heated to 300 8C, fully decomposing NaAlH4; hence these data represent
the reversible hydriding of NaH1Al to Na3AlH6. The experiment time
constants t are given in the legend along with the maximum total data
acquisition times ot given in parentheses for each P–C isotherm data set.
The theoretical maximum capacity of 1.85 wt.% per Eq. (2) is shown as
the dashed line. The data for the ball milled samples have been scaled to
correct for NaAlH4 phase dilution by the 20 wt.% diamond (Dia. milled)
and the 20 wt.% Al and 10 wt.% diamond (Dia.1Al milled).
One might assume that ball milling the NaAlH4 might

262
G.P. Meisner et al. / Journal of Alloys and Compounds 337 (2002) 254–263
simply promote initial dehydriding and subsequent rehy-
driding by facilitating hydrogen transport between the bulk
and the surrounding atmosphere. However, the as-received
material has a grain size of 120 mm and dehydrides faster
than the purified material with a grain size of 12 mm.
Moreover, after the first heating to 300 8C, our micro-
graphs show that the morphologies of all three of these
materials are roughly similar.
Despite the fact that all of the materials tested here
showed considerable hysteresis in their P–C isotherms,
and thus do not represent chemical equilibria, it is still
interesting to compare our data with equilibrium data. A
Van’t Hoff-type plot may be constructed from data for the
diamond ball milled, purified, Ti-doped, and diamond1Al
ball milled materials by plotting the midpoints of the
adsorption and desorption plateaus as a function of 1/T.
Fig. 11 exhibits this data and compares it with the
of NaH1Al to Na3AlH6, probably because neither the
pressurizing nor the depressurizing plateaus are near
equilibrium in our experiments. On the other hand, our
value for DH is comparable to the calorimetric result of
Claudy et al. [13] who find DH541.4 kJ/mol Na3AlH6 for
Na3AlH6 →3NaH1Al13/2H2, or 27.6 kJ/mol H2. There
is considerable variability in values for DH reported in the
literature and this may be due to the slow kinetics of the
hydriding and dehydriding reactions as we have observed.
Even more extreme kinetic limitations are associated with
dehydriding, as the slope of the experimental curve for
diamond milled and pure materials are both near 144
kJ/mol H2. Although the slopes are all much steeper than
the hydriding curve, the dehydriding data show wide
variations among our three types of samples.
´
measurements of Bogdanovic et al. [12] for the NaH to
5. Conclusion
Na3AlH6 transition in Ti-doped material.
´
Bogdanovic et al. [12] attributed the nearly equal
These diverse experiments form a picture of purified
NaAlH4 which is strongly stabilized against decomposi-
tion. As-received impure material, superficially Pt-doped
purified material, Ti-doped purified material, or even ball
milled as-received material all decompose at lower tem-
peratures to release hydrogen. Even after these materials
have been heated to over 300 8C, completely converting
them to NaH1Al, each product may be at least partially
rehydrided to Na3AlH6 at pressures below 4 MPa. The
purified material only reaches 70% of its theoretical
maximum capacity after a 181 h P–C isotherm experi-
ment, while the Ti-doped material reaches full capacity
during an 18 h experiment. Despite the fact that the ball
milled as-received material is initially partially decom-
posed, it may be nearly fully reconstituted to Na3AlH6
during an 18 h P–C isotherm experiment.
plateau pressures observed for the sorption and desorption
limbs of their P–C isotherms to the extraordinarily long
equilibration time in their experiments on Ti-doped materi-
al, 0.5–6 days per data point. By contrast, our sorption
plateau pressures are always observed to be much larger
than the desorption pressures. A long equilibration time
might be expected for a hydriding reaction as complex as
Eq. (2), where for each unit cell of Na3AlH4 formed from
3NaH1Al, the hydriding reaction must move an Al and
three Na atoms into proper position. Surprisingly, the
hydriding data (open symbols in Fig. 11) for the purified,
diamond ball milled, and Ti-doped materials are quite
similar, forming a data set consistent with a constant slope
of 22.7 kJ/mol H2. Even the P–C isotherm data for the
purified sample (filled circles in Fig. 10), which hydrided
to only 30% of the hydrogen sorption of the Ti-doped or
diamond milled sample, coincided with the general hydrid-
ing ‘equilibrium’ curve (open symbols in Fig. 11) to within
experimental error. Our 22.7 kJ/mol H2 value for the
enthalpy, DH, is not in particularly good agreement with
Acknowledgements
Thanks to Jerry Moleski and Martin Meyer for deft
experimental aid, to Curt Wong for the scanning electron
micrographs, to Bouziane Yebka for contributions to this
work, and to Jan Herbst for many helpful suggestions.
´
the 47 kJ/mol H2 measurement by Bogdanovic et al. [12]
also obtained from Van’t Hoff-type plots1 for the hydriding
1To estimate the enthalpy of hydrogen sorption, we use the Clapeyron
equation dP/dT5DH/TDV, which gives DH in the same molar units as
DV, i.e. if DV is the molar volume of an ideal gas DV5RT/P (R is the
molar gas constant) then DH is the enthalpy per mole of gas. This
assumes the volume of the condensed phase (solid or liquid hydride
phases) can be neglected compared to the volume of the gas. We then
have dP/P5DH/R (dT/T 2), and finally d(ln P)5(2DH/R)d(1/T) which
is the Clausius–Clapeyron equation. Thus, analysis of equilibrium plateau
pressures as a function of temperature (Van’t Hoff plot of ln P versus
1/T) can give an estimate of the enthalpy of hydrogen sorption DH5
2R[d(ln P)/d(1/T)] in units of energy per mol of hydrogen gas sorbed.
To compare to results determined calorimetrically, one must bear in mind
that 1 mol of NaAlH4 desorbs 1 mol of H2 in decomposing to 1/3 mol of
Na3AlH6 (plus 2/3 mol of Al), and desorbs an additional 1/2 mol of H2
in the final decomposition step into 1 mol of NaH (plus 1 mol of Al).
References
[1] P. Fairley, Technol. Rev. November/December (2000) 54.
[2] M. DeLuch, Hydrogen Fuel Cell Vehicles, Institute of Transporta-
tion Studies, UC Davis, Davis, CA, 1992.
[3] G.G. Tibbetts, G.P. Meisner, C.H. Olk, Carbon 39 (2001) 2291.
[4] C.C. Ahn, Y. Ye, B.V. Ratnakumar, C. Witham, R.C. Bowman Jr., B.
Fultz, Appl. Phys. Lett. 73 (1998) 3378.
[5] M.S. Dresselhaus, K.A. Williams, P.C. Eklund, MRS Bull. 24 (11)
(1999) 45.
´
[6] B. Bogdanovic, M. Schwickardi, J. Alloys Comp. 253 (1997) 1.

G.P. Meisner et al. / Journal of Alloys and Compounds 337 (2002) 254–263
263
[7] R.A. Zidan, S. Takara, A.G. Hee, C.M. Jensen, J. Alloys Comp. 285
(1999) 119.
[11] K.J. Gross, S. Guthie, S. Takara, G. Thomas, J. Alloys Comp. 297
(2000) 270.
¨
[8] A. Zaluska, L. Zaluski, J.O. Strom-Olsen, J. Alloys Comp. 298
´
[12] B. Bogdanovic, R.A. Brand, A. Marjanovic, M. Schwickardi, J.
(2000) 125.
¨
Tolle, J. Alloys Comp. 302 (2000) 36.
[9] C.M. Jensen, K.J. Gross, Appl. Phys. A 72 (2001) 213.
[10] D.F. Shriver, The Manipulation of Air-Sensitive Compounds, Wiley,
New York, 1986.
[13] P. Claudy, B. Bonnetot, G. Chahine, J.M. Letoffe, Thermochim.
Acta 38 (1980) 75.