Phase changes and hydrogen release during decomposition of sodium alanates

Article abstract of DOI:10.1016/S0925-8388(02)00996-9

We have constructed a hermetically sealed high-temperature cell for an X-ray diffractometer which allows in situ identification of structural phase changes during the several thermal decomposition stages of sodium alanate materials. Comparing X-ray data with thermogravimetry measurements allows identification of phase changes that correlate with hydrogen release. Measurements made while uniformly ramping up the temperature of purified NaAlH₄ indicate that all of the hydrogen release corresponds to the formation of Al or the formation and decomposition of the Na₃AlH₆ phase. The melting of the NaAlH₄ phase does not correlate well with sample mass loss in the purified material. For as-received technically pure NaAlH₄, hydrogen release lags both Al formation and the formation and decomposition of the Na₃AlH₆ phase. In ball-milled as-received NaAlH₄ (ball-milled with diamond powder to achieve ～2 μm crystallite size) there is good correspondence between hydrogen release and either Al formation or the formation and decomposition of the Na₃AlH₆ phase, except that hydrogen release lags the phase changes by a few degrees. For purified NaAlH₄ ball-milled in either a steel or tungsten carbide jar, there is a much smaller lag in hydrogen release. For Ti doped material, the phase changes to NaH are completed at the lowest temperature observed in any of the samples, 215°C, but the TGA measurements are not reliable because of mass loss from decomposition of the catalyst butoxide anion. Among these samples, the delay in hydrogen release does not correlate well with crystallite size; indicating that the delay is not due to hydrogen diffusion, but the delay is definitely decreased by ball milling or adding catalyst material or even impurities.

Full text of DOI:10.1016/S0925-8388(02)00996-9

Journal of Alloys and Compounds 350 (2003) 136–144
L
www.elsevier.com/locate/jallcom
Phase changes and hydrogen release during decomposition
of sodium alanates
b
b
b
b
M.P. Balogha , , G.G. Tibbetts , F.E. Pinkerton , G.P. Meisner , C.H. Olk
*
aGeneral Motors Research and Development Center, Chemical & Environmental Sciences Laboratory, MC 480-106-320, Warren, MI 48090-9055,
USA
bGeneral Motors Research and Development Center, Materials & Processes Laboratory, MC 480-224-320, Warren, MI 48090-9055, USA
Received 12 June 2002; received in revised form 26 August 2002; accepted 27 August 2002
Abstract
We have constructed a hermetically sealed high-temperature cell for an X-ray diffractometer which allows in situ identification of
structural phase changes during the several thermal decomposition stages of sodium alanate materials. Comparing X-ray data with
thermogravimetry measurements allows identification of phase changes that correlate with hydrogen release. Measurements made while
uniformly ramping up the temperature of purified NaAlH4 indicate that all of the hydrogen release corresponds to the formation of Al or
the formation and decomposition of the Na3AlH6 phase. The melting of the NaAlH4 phase does not correlate well with sample mass loss
in the purified material. For as-received technically pure NaAlH4, hydrogen release lags both Al formation and the formation and
decomposition of the Na3AlH6 phase. In ball-milled as-received NaAlH4 (ball-milled with diamond powder to achieve |2 mm crystallite
size) there is good correspondence between hydrogen release and either Al formation or the formation and decomposition of the Na3AlH6
phase, except that hydrogen release lags the phase changes by a few degrees. For purified NaAlH4 ball-milled in either a steel or tungsten
carbide jar, there is a much smaller lag in hydrogen release. For Ti doped material, the phase changes to NaH are completed at the lowest
temperature observed in any of the samples, 215 8C, but the TGA measurements are not reliable because of mass loss from decomposition
of the catalyst butoxide anion. Among these samples, the delay in hydrogen release does not correlate well with crystallite size; indicating
that the delay is not due to hydrogen diffusion, but the delay is definitely decreased by ball milling or adding catalyst material or even
impurities.
2002 Elsevier Science B.V. All rights reserved.
Keywords: Metal hydrides; Complex hydrides; Hydrogen storage; NaAlH4; Na3AlH6
1. Introduction
wt% hydrogen at room temperature [3,4]. We do not regard
this as a promising avenue to explore.
The automotive industry is now making substantial
investments in fuel cell development [1] in the hope that
this technology will provide efficient ground transportation
as petroleum supplies decrease in the future. In order to
avoid the major problems of refining a hydrocarbon fuel in
the vehicle, fuel cell vehicles will require a method of
storing hydrogen on board. For efficient operation and
reasonable range, it has been projected that the storage
material should be able to hold at least 6.5 wt% hydrogen
[2] and release its sorbed hydrogen below 150 8C.
There has been, however, good progress recently in
improving the hydrogen storage capacity, sorption kinetics,
and temperature behavior of sodium alanates. These ma-
terials are inexpensive and can almost meet the DOE 6.5
wt% capacity target [2]. NaAlH4 has been known for many
years for its two-step thermal decomposition [5],
NaAlH4↔1/3Na3AlH6 1 2/3Al 1 H2 3.70 wt% H2 (1)
1/3Na3AlH6 1 2/3Al 1 H2↔NaH 1 Al 1 3/2H2
Our recent studies of carbon nanofibers have convinced
us that the best of these materials will store less than 1
1.85 wt% H2,
(2)
providing a total of 5.55 wt% hydrogen at the NaH
endpoint. Further decomposition of NaH is not feasible for
automotive applications because temperatures over 300 8C
are required. Even so, the major drawback in alanate
*
Corresponding author.
E-mail address: michael.p.balogh@gm.com (M.P. Balogh).
0925-8388/02/$ – see front matter
PII: S0925-8388(02)00996-9
2002 Elsevier Science B.V. All rights reserved.

M.P. Balogh et al. / Journal of Alloys and Compounds 350 (2003) 136–144
137
decomposition according to Eqs. (1) and (2) is that
complete decomposition requires temperatures up to
300 8C, while the reverse reaction of Eq. (1) requires very
high pressures.
ball-milled as-received NaAlH4 material contained 20 wt%
of 6 mm diamond powder and was milled continuously for
1 h. The ball-milled purified NaAlH4 material was fabri-
cated similarly, but to ensure that heating and partial
desorption of the hydrogen was minimized, it was milled
in 5-min intervals followed by 30 min cooling for a total
milling time of |3 h. The bulk of the purified ball-milled
NaAlH4 material remained cooler than 100 8C although
impact point temperatures were probably considerably
higher. After we discovered substantial Fe in the samples
ball-milled in the steel jar, we ball-milled a sample of pure
NaAlH4 (no diamond) in a tungsten carbide jar using WC
balls and the above 5-min interval techniques for com-
parison.
Interest in these materials was reawakened by a 1997
´
paper by Bogdanovic and Schwickardi [6], which showed
how reversibility could be catalyzed with transition metals,
particularly Ti. Zidan, Takara, Hee and Jensen [7] then
reported that doping NaAlH4 with both Ti and Zr lowered
the dehydriding temperatures of both stages of decomposi-
tion, offering measurable dehydriding below 100 8C.
¨
Zaluska, Zaluski, and Strom-Olsen [8] succeeded in ac-
celerating the dehydrogenation kinetics by ball-milling
alanates both without additions and with carbon.
Here we seek to understand and compare these improve-
ments by correlating phase changes due to heating mea-
sured by X-ray diffraction in purified, as-received, ball-
milled, and Ti-doped NaAlH4 materials with the hydrogen
release temperatures measured by thermogravimetric anal-
ysis.
2.2. Procedures
Thermogravimetric analysis (TGA) was performed using
a Perkin-Elmer System 7 thermogravimetric analyzer. The
apparatus was housed in an Ar glove box to protect the
air-sensitive samples during loading. Samples ranging
from 5 to 14 mg in weight were placed in either a stainless
steel or a Pt pan and heated at 1 8C/min from 30 to 300 8C
in a 45-sccm flow of Ar.
2. Experiment
2.1. Materials
Real-time in-situ X-ray (XRD) diffraction data were
collected using Cu Ka radiation in a Bruker AXS General
Area Detector Diffractometer System (GADDS). A special
sample cell and furnace, shown in Fig. 1, were designed to
minimize oxidation and moisture contamination of these
deliquescent materials throughout the diffraction experi-
ment. An XRD capillary tube seals to the brass base with
O-rings, and a small passageway connects the capillary
tube to the pressure manifold via a quick disconnect fitting.
The as-received NaAlH4 material was purchased from
Lancaster Synthesis, Inc. and has a nominal technical
purity of 90 wt% with Al and Na3AlH6 as the major
impurities.
The purified NaAlH4 was synthesized in our laboratory
from a 1.0-M solution of commercial NaAlH4 in THF
(tetrahydrofuran) prepared using the Shlenk method [9].
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 solu-
tion. One-hundred millilitres 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 another 2 h and filtered. The residue was washed
with toluene three times, and then filtered and dried under
vacuum overnight at room temperature. After drying, the
purified NaAlH4 was obtained as a fine white powder.
The Ti-doped NaAlH4 material 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.
Steel 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 21 g for a 2-g
typical sample size. Steel ball-milling was performed in an
O-ring sealed steel jar under 1 atmosphere of Ar. The steel
Fig. 1. X-ray diffraction apparatus and scattered X-ray collection geome-
try used to acquire real-time in situ phase data.

138
M.P. Balogh et al. / Journal of Alloys and Compounds 350 (2003) 136–144
Table 1
X-Ray diffraction parameters used to monitor sample composition
Phase
JCPDF number
Miller indices
Range (2u )
NaAlH4
Na3AlH6
Al
NaH
Diamond
Background
73-0088
20-1072
4-787
2-809
6-675
–
(112)
(020)(112)(200)
(111)
(111)
(111)
–
28.6–30.2
31.8–33.2
37.4–38.9
30.5–31.8
43.6–44.6
24–25
Two posts guide and align the capillary tube in a miniature
tube furnace which has two conical windows: one permits
the primary X-ray beam access to the sample and the other
permits detection of the diffracted and scattered X-ray
beams by the GADDS. The sample cell and furnace can be
heated to 500 8C, and the temperature is monitored by a
thermocouple adjacent to the sample. The temperature at
the window of the furnace is about 5% cooler than the
furnace block, but the sample temperature could be
calibrated using melting point standards to an absolute
accuracy of 1%.
Prior to the diffraction measurements, we loaded and
sealed the various samples into 1–1.5 mm quartz capillary
tubes while inside an inert atmosphere glove box. The
XRD data were then collected using a 0.8 mm collimator
and a detector distance of 15 cm, with the center of the
sample cell at angles V5208 and 2u5358. This allowed
collection of integrated diffraction data for angles 17.48#
2u #52.38. The XRD experiments were performed while
heating the various NaAlH4 samples from room tempera-
ture to above 400 8C at 2 8C/min in 1 bar of H2.
Diffraction patterns were collected every 30 s, and 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 due
to the relatively large crystallite size compared to the
capillary diameter. The compositions of the different
samples were monitored from the integrated net intensity
of the prominent Bragg reflection for each phase, as listed
in Table 1. For Na3AlH6, the 2u apertures are sufficiently
large so that phase information on both the a and b phases
is collected, but here we make no distinction between the
two. Non-crystalline material was monitored by measuring
the background intensity. In the case of the ball-milled
material, an elevated background was observed and is
believed to be a result of fluorescence of iron contaminant
introduced during ball-milling.
Fig. 2. X-ray diffraction intensities of phases identified during the heating
of purified NaAlH4 compared with observed (TGA Expt) and calculated
TGA curves based on either the Al (Al calc) or the Na3AlH6 (Na calc)
intensities.
ture to over 300 8C. We observe the clear signature of a
melting transition near 180 8C by the disappearance of the
diffraction peaks of the NaAlH4. Immediately above this
temperature there is a rapid rise in background scattering
that we associate with the liquid NaAlH4 phase. Above
210 8C, there is a gradual increase in both the cubic Al
reflections and the Na3AlH6 reflections. Near 240 8C the
Na3AlH6 concentration reaches its maximum and simul-
taneously NaH begins forming. NaH then increases as
Na3AlH6 decreases, reaching its saturation value near
280 8C.
Superimposed on this plot is the 1 8C/min observed
TGA result for mass loss due to H2 desorption on this plot.
Surprisingly, there is no mass loss whatsoever that can be
correlated with the disappearance of the NaAlH4 solid
phase. The total mass loss seems to correspond with the
appearance and disappearance of Na3AlH6 and the
simultaneous increase in NaH.
As-received NaAlH4 shows qualitatively different be-
havior on heating (Fig. 3). First, the measurement of
NaAlH4 phase concentration shows more noise variation
as the temperature is changed because only a small number
of the 100-mm grains fit within the X-ray beam. When
these shift as temperature is raised, the X-ray diffraction
signal fluctuates considerably. The NaAlH4 phase dis-
appears at a slightly lower temperature than for purified
NaAlH4. It is replaced not only by an increase in the
background scattering, but by an Na3AlH6 phase which
reaches its maximum concentration near 190 8C, then
diminishes gradually and disappears near 255 8C. NaH
formation again correlates with Na3AlH6 disappearance;
NaH concentration becomes appreciable above 210 8C and
reaches its maximum near 260 8C. The Al signal also tells
a key story; it rises rapidly as the Na3AlH6 phase forms
near 175 8C and rises again as this phase disappears and
NaH forms.
The materials were also analyzed for metallic contami-
nation after preparation by inductively coupled plasma
atomic emission spectroscopy.
3. Results
Fig. 2 shows the relative concentrations of the phases
observed as purified NaAlH4 is heated from room tempera-
The observed TGA curve for the as-received material is

M.P. Balogh et al. / Journal of Alloys and Compounds 350 (2003) 136–144
139
Fig. 3. X-ray diffraction intensities of phases identified during the heating
of as-received NaAlH4 compared with observed (TGA Expt) and
calculated TGA curves based on either the Al (Al calc) or the Na3AlH6
(Na calc) intensities.
Fig. 5. X-ray diffraction intensities of phases identified during the heating
of steel ball-milled purified NaAlH4 compared with observed (TGA Expt)
and calculated TGA curves based on either the Al (Al calc) or the
Na3AlH6 (Na calc) intensities.
consistent with this picture. It shows a somewhat tardy but
large mass loss near 190 8C, and a second smaller mass
loss between 230 and 270 8C. This TGA result is much
more consistent with our expectations based on Eqs. (1)
and (2) than was the TGA result for purified NaAlH4.
Data for steel ball-milled as-received NaAlH4 (Fig. 4)
are qualitatively similar, but the smaller grain size of about
2 mm gives each phase more overall diffraction intensity
and hence a less noisy diffraction signal. The NaAlH4
phase disappears between 160 and 170 8C, while the
Na3AlH6 phase grows over the same interval, exactly
replacing the NaAlH4. In this sample the background
remains constant over the whole temperature range. Fur-
thermore, the Al signal is clearly above zero even before
decomposition of the NaAlH4 phase occurs. It then rises
sharply at 170 8C where the NaAlH4 phase converts to
Na3AlH6 and again as the Na3AlH6 converts to NaH. The
diamond (111) intensity trace is independent of tempera-
ture; for simplicity it is not shown on Fig. 4.
Measurements by Inductively Coupled Plasma Atomic
Emission Spectroscopy reveal that the steel ball-milled
samples gain large amounts of iron during the ball-milling
process. As Table 2 shows, the as-received NaAlH4 gains
1.9 wt% Fe during ball-milling. The inert fraction of Ti
doped NaAlH4 is listed as 0.113, the mass equivalent of 2
mole% of Ti[O(CH2)3CH3]4, even though we shall see
that the butoxide anion decomposes in the temperature
range studied here.
The TGA data for steel ball-milled as-received NaAlH4
(Fig. 4) correlate with XRD observations even better than
those for as-received NaAlH4, with the low temperature
mass loss only a few degrees above the NaAlH4 to
Na3AlH6 conversion, and a second mass loss correlating
well with the formation of NaH. The large X-ray fluores-
cence associated with the 1.9 wt% of Fe probably accounts
for the enhanced X-ray scattered background.
Fig. 4. X-ray diffraction intensities of phases identified during the heating
of steel ball-milled as-received NaAlH4 compared with observed (TGA
Expt) and calculated TGA curves based on either the Al (Al calc) or the
Na3AlH6 (Na calc) intensities.
The 3-h slow steel ball-milling of the purified NaAlH4
(Fig. 5) introduces a much larger Fe concentration, 23 wt%
(Table 2). The enhanced Fe concentration is accompanied
Table 2
Impurity concentrations in sodium alanate samples
Sample
As-received
Purified
Steel BM
as-received
Steel BM
purified
WC BM
purified
Ti-doped
Diamond (wt%)
Fe (wt%)
f (fraction inert)
0
0
0
0
0
0
20
1.9
0.219
20
23
0.43
0
0
0
0
0
0.113

140
M.P. Balogh et al. / Journal of Alloys and Compounds 350 (2003) 136–144
until Na3AlH6 begins to disappear at 200 8C. All phase
changes are complete before 230 8C. The TGA data
correspond very nicely with the phase change data.
To determine the behavior of pure NaAlH4 ball-milled
with as little impurity as possible, a sample of pure
NaAlH4 was ball-milled using a tungsten carbide jar and
balls and no diamond addition using the 5-min on (3-h
total) procedure. XRD and TGA data for this sample are
shown in Fig. 6. The sample shows clear evidence of
Na3AlH6 at 150 8C, but Na3AlH6 does not peak until
228 8C. There is no trace of Al at lowest temperatures, but
Al begins growing as NaAlH4 starts to disappear near
170 8C, reaching its maximum near 245 8C. Thus, in the
temperature required for Na3AlH6, NaH, and Al formation
the XRD behavior of this sample is intermediate between
the purified material and the steel ball-milled purified
sample. Moreover, only in this sample do we observe
NaAlH4 pre-decomposition manifested by Na3AlH6 for-
mation totally without Al formation. The TGA data
resembles that for the purified crystals in that there is no
mass loss corresponding to the disappearance of the
NaAlH4 phase; the entire mass loss corresponds in tem-
perature to the formation and decomposition of Na3AlH6.
Pure NaAlH4 doped with 2 mol% of Ti[O(CH2)3CH3]4
was also studied by these methods. Fig. 7 shows that many
of the phase changes occur at the lowest temperatures
observed. Solid NaAlH4 disappears completely to form
Na3AlH6 near 170 8C, and the latter decomposes com-
pletely before 220 8C. NaH begins forming at the earliest
observed temperature, 175 8C, and reaches its maximum
value at 215 8C. However, the TGA results do not clearly
track these observations, as mass does not begin dropping
until over 180 8C. After an initial drop which is clearly at a
Fig. 6. X-ray diffraction intensities of phases identified during the heating
of purified NaAlH4 ball-milled in a WC jar compared with observed
(TGA Expt) and calculated TGA curves based on either the Al (Al calc)
or the Na3AlH6 (Na calc) intensities.
by an increase in the background of scattered X-rays and a
loss of X-ray penetration which considerably degrades the
signal/noise ratio of the data. Nevertheless, in contrast
with the steel ball-milled as-received material, the steel
ball-milled purified material does not show perceptible
decomposition of the original NaAlH4; neither Al nor
Na3AlH6 is visible at below 140 8C. However, as Fig. 5
shows, the steel ball-milled purified material decomposes
in a mode very similar to that of the steel ball-milled
as-received material: NaAlH4 converts to Na3AlH6 above
160 8C. Na3AlH6 begins growing at 140 8C, but this
constituent reaches its maximum near 200 8C, and dis-
appears near 220 8C. Al begins growing as the NaAlH4
disappears at 160 8C, while NaH does not begin to appear
Fig. 7. X-ray diffraction intensities of phases identified during the heating of 2% Ti-doped NaAlH4 compared with observed (TGA Expt) and calculated
TGA curves based on either the Al (Al calc) or the Na3AlH6 (Na calc) intensities.

M.P. Balogh et al. / Journal of Alloys and Compounds 350 (2003) 136–144
141
temperature above that for conversion of NaAlH4 to
Na3AlH6, mass is lost at a slow and nearly continuous rate
up to 300 8C and beyond. Since this mass loss ultimately
extends to more than the 10.6 wt% total H2 present in the
sample, it must correspond to not only the decomposition
of the alanates but also the loss of the butoxide anion. This
TGA was performed several times with both a Pt pan and a
stainless steel pan, with and without a lid, with only minor
differences in an effort to detect Na volatilization.
generated by the following formula from the Na3AlH6(T)
data:
Al1 2 Al0
]
Nacalc(T) 5 100 2
F G
Al1
Na3AlH6(T)
(Na3AlH6)1 2Na3AlH6(T)
]
]
3 3.7
11.85
S
D
(Na3AlH6)1
(Na3AlH6)1
3 (12f )
(4)
Here the first term in the curved brackets is evaluated
only for temperatures below that at which Na3AlH6(T)
reaches its maximum, while the final term in the curved
brackets is evaluated only after Na3AlH6(T) reaches its
maximum. The curves from Eqs. (3) and (4) show
moderately good agreement with each other and with the
observed TGA of Fig. 4. The Al calculation underpredicts
the first decomposition step by 7 8C, while the Na calcula-
tion underpredicts it by 13 8C. Both predict the second
decomposition more accurately. By construction they agree
on the total decomposition at 300 8C, yet both overestimate
only slightly the mass of hydrogen lost at that temperature.
Steel ball-milled purified NaAlH4 (Fig. 5) shows excel-
lent agreement between the two mass loss calculations and
the experimental TGA, especially considering the poor
signal quality of the XRD data. The calculations do not
appear to lag the TGA features.
4. Calculated TGA curves
Let us attempt to generate calculated TGA curves from
phase changes observed by XRD for quantitative com-
parison with the observed TGA data. Steel ball-milled
as-received NaAlH4 (Fig. 4) shows the most encouraging
agreement. The first calculated TGA curve is based on the
observed X-ray concentration of Al and the suppositions
that:
1. The Al present below 170 8C, Al0, corresponds to initial
decomposition of the NaAlH4 phase after fabrication or
during ball-milling.
2. The Al concentration observed above 250 8C, Al1,
corresponds to the complete decomposition of all
hydride phases except NaH.
Fig. 3 shows similar calculations made for as-received
NaAlH4. The decomposition is qualitatively similar to steel
ball-milled NaAlH4, both as-received (Fig. 4) and purified
(Fig. 5), except that the calculations now lag the TGA
trace by 20 8C.
3. Hydrogen (5.55 wt%) is released during the complete
formation of Al from the pure alanate. However, a
fraction ‘f’ of inert material, such as abrasive or
catalyst, may be mixed with the sample and lower this
number.
In contrast to the above three materials (Figs. 3–5),
purified NaAlH4 (Fig. 2) has a distinctly different TGA
curve. Even so, the two calculations still give a qualitative
approximation to the observed TGA curve, even though
they underpredict the desorption temperature by 27 8C or
so. For purified NaAlH4, the observed TGA curve agrees
with the calculation in that it shows a mass decrease of the
full 5.55%. In this case there is neither Al nor inert matter
visible in the low temperature XRD data, so there is no
correction of the calculated maximum value. The Na
calculation is rather noisy because of the large crystallite
size of the sample.
For the purified WC ball-milled NaAlH4, (Fig. 6)
calculations based on the Na3AlH6 signal suffer an imme-
diate 0.5 wt% offset from the experimental TGA signal
because the Na3AlH6 signal used in Eq. (4) is appreciable
at 150 8C. With the exception of this offset and the
somewhat noisy Na3AlH6 signal, however, agreement
between the two calculations and the measured TGA is
quite good until 245 8C. However, since the Na3AlH6
concentration continues to fall and the Al concentration
continues to rise above this temperature, both calculated
curves continue to decline while the experimental TGA
curve flattens out. Moreover, the two calculated mass loss
The curve in Fig. 4 labeled ‘Al calc’ is therefore
generated by the following formula from the XRD data
Al(T):
Al(T) 2 Al0
]
Alcalc(T) 5 100 2 5.55
(1 2 f )
(3)
Al1
A second and partially independent calculated curve
may be generated from the formation and disappearance of
the Na3AlH6 through the suppositions that:
1. Decomposition of the NaAlH4 phase is accompanied by
the release of 3.70 wt% hydrogen as Na3AlH6 rises
from zero to its maximum value (Na3AlH6)1.
2. Decomposition of the Na3AlH6 phase is accompanied
by the release of 1.85 wt% hydrogen as its concen-
tration drops from its maximum to zero.
3. Both of the above phases must be proportionately
reduced to account for the fraction Al0 /Al1 of all
hydride decomposed before heating and the inert frac-
tion f.
The curve in Fig. 4 labeled ‘Na calc’ is therefore

142
M.P. Balogh et al. / Journal of Alloys and Compounds 350 (2003) 136–144
curves predict a mass loss of 5.55 wt%, while the observed
TGA has a maximum weight loss of 4.2 wt%.
ranges from good to acceptable. For purified NaAlH4,
because the observed TGA curve shows hydrogen release
very close to the stoichiometric 5.55 wt%, the model by
construction fits the high-temperature endpoint closely. For
both iron ball-milled materials, the TGA experiment is
overestimated only slightly by the two calculations. For
as-received NaAlH4, the model overestimates the high-
temperature TGA weight loss by about 10%, which
corresponds to the nominal purity of the material. For WC
ball-milled pure material the TGA decomposition is some-
what less than that obtained from the two calculations.
This sample is anomalous in that the absence of any low
temperature Al signal indicates that ball-milling did not
decompose a measurable amount of NaAlH4, while the
appreciable low temperature Na3AlH6 signal indicates that
about 16% of the sample has already converted to
Na3AlH6. This explains why the Na calc curve starts at
somewhat below 100%.
The calculated TGA curves for the Ti-doped NaAlH4
(Fig. 7) are the least satisfactory of the materials tested.
Although the TGA curves calculated by both the Na and
Al method agree very nicely with each other, neither
agrees at all well with the observed TGA curve, lagging
the first decomposition by 17 8C. The observed TGA curve
does not show any feature which might be associated with
the second decomposition; rather, it seems to continuously
decrease until well above 300 8C. It seems clear that the
observed TGA curve above 230 8C shows some serious
sample decomposition which strongly interferes with our
ability to calculate the TGA from the XRD phase in-
formation. This most likely results from the decomposition
of the butoxide catalyst anions.
5. Discussion
Because the XRD data detects only Al crystallites of
appreciable lattice coherence length, and cannot in princi-
ple measure dissolved Al or very small aggregates of Al
atoms, the number Al0 which we determine from the low
temperature Al X-ray signal is an underestimate of all
initial free Al. Therefore, Eqs. (3) and (4) should, as they
do, overestimate the total decomposition which the TGA
experiment measures. This overestimate is most extreme in
the as-received and WC ball-milled pure NaAlH4, indicat-
ing that many Al atoms present in this sample are
dispersed too broadly to scatter X-rays and appear as XRD
data. These Al atoms presumably come from very slow
decomposition of the NaAlH4 at room temperature, and
cannot easily aggregate at low temperatures. This is also
the likely explanation for the absence of Al observed at
150 8C in the WC ball-milled purified NaAlH4 where the
Na3AlH6 shows a signal appreciable above 0. Proper
normalization of the calculated TGA traces would move
them closer to the experimental curve.
Thus, the decomposition of as-received NaAlH4 and all
three ball-milled NaAlH4 materials seems to be consistent
with Eqs. (1) and (2) above, with hydrogen release
correlating to a certain extent with the formation and
decomposition of Na3AlH6. The formation of Na3AlH6
can occur at various temperatures for materials prepared by
different methods, but it is always in better agreement with
the formation of Al crystallites. This seems to be in
agreement with the suggestion of Gross et al. [11] that
hydrogen is primarily bonded as AlH3. They suggested
recasting Eqs. (1) and (2) as
Comparison of Figs. 2 and 7 shows a decrease of the
temperature at which the alanate completely converts to
NaH of 60 8C in the presence of the Ti catalyst. This is in
´
excellent agreement with results observed by Bogdanovic
and Schwickardi [6] and Jensen, Zidan, Mariels, Hee and
Hagen [10]. A similar decrease due to ball-milling of the
pure material in Fe crucibles is in agreement with the more
qualitative data presented by Zaluska, Zaluski, and Strom-
Olsen [8].
¨
A recent paper by Gross, Guthrie, Takara, and Thomas
[11] utilized somewhat different equipment to obtain X-ray
diffraction phase identification data on purified and cata-
lyzed NaAlH4. Our results differ with theirs in some
details. Gross et al. [11] observed rapid and reversible
changes in the NaAlH4 peak intensities upon warming, and
attributed them to distortions in the crystal lattice. Exami-
nation of our figures shows very large NaAlH4 peak
intensity fluctuations in Fig. 3, lower fluctuations in Fig. 2,
and least in Fig. 4; the fluctuations decrease as crystallite
size decreases from 100 to 20 to 2 mm. We believe these
fluctuations are large when a small number of large
crystallites expand with heating and shift in and out of the
diffracting condition. With the ball-milled 2 mm material,
positional movements average out, and only small fluctua-
tions are seen.
Furthermore, we do not see any trace of either the X1 or
X2 phase observed by Gross et al. [11]. We do, however,
see traces of NaSiO3 and a complex Na–Al silicate
forming over 400 8C which we believe incorporate silicates
from the sample tube. Perhaps the suggestion of Gross et
al. that the X2 phase may incorporate Be from their
window is correct.
(NaH)(AlH3)↔1/3 (NaH)3(AlH3) 1 2/3 (AlH3)↔
(NaH) 1 (AlH3)
(6)
(7)
Let us take a somewhat deeper look at what the TGA
modeling calculation reveals about decomposition in so-
dium alanates. For all samples except the Ti catalyzed:
In general the normalization of our TGA calculations
AlH3↔Al 1 3/2H2,
which clearly have the desired form. In the purified
NaAlH4 the AlH3 seems to be much more long lived than

M.P. Balogh et al. / Journal of Alloys and Compounds 350 (2003) 136–144
143
in the other two samples, perhaps because of the absence
of catalysts or surface active sites, and therefore requires
much higher temperatures for its decomposition and the
release of hydrogen.
Na3AlH6 reaches its maximum for the different samples.
This maximum may be precisely defined from the data
and, as we have seen, approximates the mid-point in
hydrogen desorption temperature. The top panel of Fig. 8
shows that Na3AlH6 in the as-crystallized purified material
maximizes at by far the highest temperature, and that
ball-milling alone in the WC jar lowers the temperature
about half as much as ball-milling in a steel crucible with
its consequent Fe addition. Na3AlH6 in the as-received
material in its original crystallized form reaches a maxi-
mum at a much lower temperature, which, nevertheless, is
lowered still more by ball-milling in a steel jar. Na3AlH6
in the Ti doped purified material maximizes at the lowest
temperature, barely above the melting temperature of as-
received NaAlH4.
A second plot (Fig. 8, bottom) shows the temperature
difference by which the average of the two calculated
curves lead the TGA experimental curve at the midpoint of
the first discernible desorption. This ‘retardation’ measures
how much later in temperature the hydrogen leaves the
sample after XRD determines a phase change. This
retardation temperature is greatest for the samples com-
posed of large crystallites, although it does not exactly
correspond to crystallite size. It is substantially smaller for
both types of ball-milled samples.
Adding to these complications is the fact that the
delayed release of hydrogen during the thermal decomposi-
tion of alkali metal alanates is strongly affected by the
slow kinetics of decomposition. Our experiments with
LiAlH4 show that increasing the heating rate by a factor of
10, from 0.2 to 2 8C/min increases the observed decompo-
sition temperatures by about 28 8C. For a simple two-fold
increase in the ramp rate, a lag of about 9 8C is evident.
Nevertheless, the TGA temperatures are still retarded
compared to the phase change temperatures by about
10 8C. Under these conditions our choice of 1 8C/min as a
ramp rate for the TGA experiments and 2 8C/min for the
ramp rate of the XRD phase change experiments may seem
rather awkward. However, it should be noted that this
selection should make the phase change data lag the TGA
data, which is the opposite of our observations in every
experiment. Moreover, the observation that our mass loss
lags vary widely from sample to sample underscores the
fact that the temperature lag is very specifically materials
dependent.
One might hope that the delayed release of hydrogen
observed in the TGA experiments compared to tempera-
tures at which phase changes occur might be understand-
able based on the mobility of hydrogen in the alanate
lattice. Qualitatively consistent with this idea is the ob-
servation that the most finely ground purified ball-milled
material exhibits prompt hydrogen release while the as-
received ball-milled material exhibiting 10 8C delay has 2
mm crystallite size. As-received NaAlH4 with a 20 8C
delayed release has 100 mm crystallite size, as observed by
scanning electron microscopy [12]. However, purified
NaAlH4, which exhibits the largest (27 8C) delayed re-
lease, has an intermediate crystallite size of 20 mm.
Moreover, calculations based on hydrogen diffusion rates
out of a spherical sample [13], approximating the diffusion
coefficients with those for H in Pd or Ni [14], give
hydrogen equilibration times of less than 10 s even for the
100-mm crystallites. Slow hydrogen mobility through the
crystallite, thus, does not appear to account for tardy
hydrogen release. It seems likely that iron ball-milling
accelerates hydrogen release partially through Fe catalysis.
This is also consistent with the somewhat lower decompo-
sition temperatures observed for every phase of the iron
ball-milled purified material with 23 wt% Fe than for
ball-milled as-received NaAlH4 with 1.9 wt% Fe.
In conclusion, the purified material is quite difficult to
decompose: WC ball-milling can drop the decomposition
temperature somewhat, but steel ball-milling lowers the
Ti-doping appears to be similar to Fe doping in that the
decomposition of the NaAlH4 phase appears nearly com-
plete at 170 8C; it is superior to Fe doping in that NaH
phase is completely formed at 215 8C rather than 230 8C.
We will present two plots in an effort to generalize the
effect of the various sample preparation procedures dis-
cussed here. The first compares the temperature at which
Fig. 8. (Top) Temperature during a 2 8C/min temperature sweep in 1 atm
of hydrogen at which the Na3AlH6 XRD line reaches its maximum,
plotted as a function of processing parameters. (Bottom) Temperature by
which the TGA measurement of hydrogen release lags the model values
calculated from the XRD data.

144
M.P. Balogh et al. / Journal of Alloys and Compounds 350 (2003) 136–144
[3] G.G. Tibbetts, G.P. Meisner, C.H. Olk, Carbon 39 (2001) 2291.
[4] F.E. Pinkerton, M.S. Meyer, G.G. Tibbetts, Proceedings of the 11th
Canadian Hydrogen Conference, Victoria, BC, 2001.
decomposition temperature even further. Ti doping is
about the best procedure of all. Surprisingly, the as-re-
ceived 90% pure material decomposes at nearly as low a
temperature as the Ti catalyzed material, particularly after
iron ball-milling.
[5] C.M. Jensen, K.J. Gross, Appl. Phys. A 72 (2001) 213.
´
[6] B. Bogdonavic, M. Schwickardi, J. Alloys Comp. 253 (1997) 1.
[7] R.A. Zidan, S. Takara, A.G. Hee, C.M. Jensen, J. Alloys Comp. 285
(1999) 119.
¨
[8] A. Zaluska, L. Zaluski, J.O. Strom-Olsen, J. Alloys Comp. 298
(2000) 125.
Acknowledgements
[9] D.F. Shriver, The Manipulation of Air-Sensitive Compounds, John
Wiley & Sons, New York, 1986.
Many thanks to Dr. Bouziane Yebka for providing the
purified and Ti catalyzed NaAlH4 materials and for useful
discussions. Thanks also to Noel Potter for the inductively
coupled plasma atomic emission spectroscopy measure-
ments and to Kathleen Taylor and Jan Herbst for many
helpful suggestions.
[10] C.M. Jensen, R. Zidan, N. Mariels, A. Hee, C. Hagen, Int. J.
Hydrogen Energy 24 (1999) 461.
[11] K.J. Gross, S. Guthrie, S. Takara, G. Thomas, J. Alloys Comp. 297
(2000) 270.
[12] G.P. Meisner, G.G. Tibbetts, C.H. Olk, M.P. Balogh, Enhancing
Low Pressure Storage in Sodium Alanates, J. Alloys Comp. 337
(2002) 254.
[13] J. Crank, The Mathematics of Diffusion, in: Oxford U. Press,
London, 1975, p. 91.
[14] W. Jost, Diffusion in Solids, Liquids, and Gases, in: Academic
Press, New York, 1952, p. 175.
References
[1] P. Fairley, Technol. Rev., November/December (2000) 54.
[2] S. Hynek, W. Fuller, J. Bentley, Int. J. Hydrogen Energy 22 (1997)
601.