Kinetics of thermal decomposition of aluminium hydride: I-non-isothermal decomposition under vacuum and in inert atmosphere (argon)

Article abstract of DOI:10.1016/j.tca.2005.08.029

Recently, interest in aluminium hydride (alane) as a rocket propulsion ingredient has been renewed due to improvements in its manufacturing process and an increase in thermal stability. When alane is added to solid propellant formulations, rocket performance is enhanced and the specific impulse increases. Preliminary work was performed at AFRL on the characterization and evaluation of two alane samples. Decomposition kinetics were determined from gravimetric TGA data and volumetric vacuum thermal stability (VTS) results. Chemical analysis showed the samples had 88.30% (by weight) aluminium and 9.96% hydrogen. The average density, as measured by helium pycnometery, was 1.486 g/cc. Scanning electron microscopy showed that the particles were mostly composed of sharp edged crystallographic polyhedral such as simple cubes, cubic octahedrons and hexagonal prisms. Thermogravimetric analysis was utilized to investigate the decomposition kinetics of alane in argon atmosphere and to shed light on the mechanism of alane decomposition. Two kinetic models were successfully developed and used to propose a mechanism for the complete decomposition of alane and to predict its shelf-life during storage. Alane decomposes in two steps. The slowest (rate-determining) step is solely controlled by solid state nucleation of aluminium crystals; the fastest step is due to growth of the crystals. Thus, during decomposition, hydrogen gas is liberated and the initial polyhedral AlH₃ crystals yield a final mix of amorphous aluminium and aluminium crystals. After establishing the kinetic model, prediction calculations indicated that alane can be stored in inert atmosphere at temperatures below 10 °C for long periods of time (e.g., 15 years) without significant decomposition. After 15 years of storage, the kinetic model predicts ～0.1% decomposition, but storage at higher temperatures (e.g. 30 °C) is not recommended.

Full text of DOI:10.1016/j.tca.2005.08.029

Thermochimica Acta 439 (2005) 32–43
Kinetics of thermal decomposition of aluminium
hydride: I-non-isothermal decomposition under
ଝ
vacuum and in inert atmosphere (argon)
∗
I.M.K. Ismail , T. Hawkins
ERC Inc at Air Force Research Laboratory/PRSP, Edwards Air Force Base, CA 93523-7689, USA
*
Received 20 May 2005; received in revised form 11 August 2005; accepted 13 August 2005
Available online 7 October 2005
Abstract
Recently, interest in aluminium hydride (alane) as a rocket propulsion ingredient has been renewed due to improvements in its manufacturing
process and an increase in thermal stability. When alane is added to solid propellant formulations, rocket performance is enhanced and the specific
impulse increases. Preliminary work was performed at AFRL on the characterization and evaluation of two alane samples. Decomposition kinetics
were determined from gravimetric TGA data and volumetric vacuum thermal stability (VTS) results. Chemical analysis showed the samples had
8
8.30% (by weight) aluminium and 9.96% hydrogen. The average density, as measured by helium pycnometery, was 1.486 g/cc. Scanning electron
microscopy showed that the particles were mostly composed of sharp edged crystallographic polyhedral such as simple cubes, cubic octahedrons
and hexagonal prisms.
Thermogravimetricanalysiswasutilizedtoinvestigatethedecompositionkineticsofalaneinargonatmosphereandtoshedlightonthemechanism
of alane decomposition. Two kinetic models were successfully developed and used to propose a mechanism for the complete decomposition of
alane and to predict its shelf-life during storage. Alane decomposes in two steps. The slowest (rate-determining) step is solely controlled by solid
state nucleation of aluminium crystals; the fastest step is due to growth of the crystals. Thus, during decomposition, hydrogen gas is liberated
and the initial polyhedral AlH
3
crystals yield a final mix of amorphous aluminium and aluminium crystals. After establishing the kinetic model,
◦
prediction calculations indicated that alane can be stored in inert atmosphere at temperatures below 10 C for long periods of time (e.g., 15 years)
without significant decomposition. After 15 years of storage, the kinetic model predicts ∼0.1% decomposition, but storage at higher temperatures
◦
(
e.g. 30 C) is not recommended.
©
2005 Elsevier B.V. All rights reserved.
Keywords: Aluminium hydride; Decomposition; Shelf-life prediction; Model free kinetics; Thermal stability
1
. Introduction
early 1970s, most of the development work in the USA came to
an end.
Alane, AlH3, is a solid rocket fuel containing (theoretically)
0.08% by weight hydrogen. It was evaluated in the early 1960s
for military applications and was formulated in propellants.
This process accelerated the development of cracks in the solid
propellant. Rocket motor firing was successfully completed on
Recently, the interest in alane has been renewed presumably
after the development of new methods of preparation and after
the discovery of new stabilizers that can slow the rate of alane
decomposition and thus increase its shelf-life. With the concern
about alane thermal stability, extensive physical and chemical
characterization of newly produced alane is needed, especially
in the area of thermal stability and hydrogen generation during
long term storage.
1
3
00-lb motors. However, the low thermal stability of the alane
used at that time resulted in the generation of excessive amounts
of hydrogen gas in the solid propellant during storage. By the
There have been a number of publications in the open liter-
ature on preparation, characterization, decomposition kinetics
and vacuum thermal stability (VTS) of alane. Depending on the
method and conditions of preparation (mainly reaction times
and temperatures), alane can be produced in at least six differ-
ଝ
Approved for public release; distribution unlimited.
∗
Corresponding author. Tel.: +1 661 275 5057; fax: +1 661 275 5435.
E-mail address: ismail.ismail@edwards.af.mil (I.M.K. Ismail).
0
040-6031/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.tca.2005.08.029

I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
33
ent crystal forms [1]. The most stable form is ␣-alane which
has been used in the current investigation. Other forms are ␤-
alane and ␥-alane which are meta-stable phases. Brower et al.
region, α = 0.6–0.9, was defined as the deceleration region, the
−1
kinetic rate was controlled by a third equation: ln(1 − α) = k3t.
The activation energies for the three regions were 97, 108 and
112 kJ/mol, respectively, within an error of 20% [6]. A compar-
ison between decomposition kinetics of AlH3 and AlD3 showed
that the deuteride was decomposing at a slower rate than the
[1] reported that in the final steps of aluminium hydride synthe-
sis, the ␤ and ␥ forms were first formed; upon further heating,
they were converted to ␣-alane. Polarizing-light microscopy
showed the crystals of the ␣-phase were mostly hexagonal and
cubical. The ␥-phase appeared as bundles of fused needles in
◦
hydride. For example, at 107 C, the induction period for the
deuteride was five times longer than that of the hydride, and
the rate constant for the deuteride (presumably in the second
region) was half of that for the hydride [6]. The differences
were attributed to the presence of different impurities and crys-
tal defects in each sample and also to differences in particle
ꢀ
random orientation. Another phase, ␥ -alane, appeared in the
examination as small multiple needles growing from one cen-
ter and forming fuzzy balls [1]. From VTS experiments [1],
determined by the Taliani method [2], the ␣-phase was the most
stable form; typically its crystals (50–150 ␮m) decomposed
◦
size. The effect of particle size of the hydride at 107 C was
◦
between 0.5 and 2.0% after 6 days at 60 C depending on particle
studied between 50 and 150 ␮m; decomposition kinetics were
highly dependent on particle size [6]. For example, to achieve
the complete liberation of hydrogen, it took ∼350 min for the
sizes >150 ␮m. but only ∼90 min for the 50 ␮m particles. The
implication of this is that the thermal stability of alane can be
enhanced, to a point, simply by increasing the starting particle
size of alane.
size.
The heat of decomposition of aluminium hydride samples
was measured experimentally in nitrogen atmosphere using a
modified bomb calorimeter accommodating a small suspended
heating oven containing the sample [3]. At 298 K, the calculated
average enthalpy of formation was −11.4 ± 0.8 kJ/mol, absolute
◦
entropywas30.0 ± 0.4 kJ/mol CandGibbsenergyofformation
Over the past decade, significant improvement was made
to produce kinetic modeling software that helped investigators
to better understand the kinetics and mechanism of chemical
reactions [7–19]. These programs proved to be useful in many
applications. Typical commercial software is available, such as
“Model Free Kinetics” by METTLER or “Thermokinetics” by
NETZSCH (which was used for the work cited in this article) and
AKTS-TA-Software. Basically these programs perform sophis-
ticated computation in a logical order. For example, a set of
experiments can be performed on the sample at different heat-
ing rates (non-isothermal method). The iso-conversion (i.e., at
constant values of α) method can normally be used to start the
analysis;ityieldsinformationonhowtheactivationenergyofthe
entire process (or its individual steps) and the pre-exponential
factor(s) could be changing as the reaction progresses; that is,
from α = 0 to 1.00. With these results taken into consideration,
the programs determine numerically the best kinetic model that
fits the data mathematically. The thermokinetics program has
several mathematical models that prescribe chemical reactions
and solid state transformations. The models include nth order
reactions, Avrami–Erofeev solid state transformation [20] and
autocatalytic reactions. A single one-step model or a combina-
tion of models (more than one-step reaction) can be chosen for
performing the analysis. These software programs have been
used in the recent literature to investigate the thermal decom-
position of many energetic materials including HMX [10–12],
ammonium perchlorate [13], ammonium nitrate [14] and ammo-
nium dinitramide [15–16]. The programs were also used to study
dimerization of cyclopentadiene [17], to predict the thermal
response of hazardous materials during storage [18] and the
thermal aging of rocket motor propellants [19]. This type of
analysis yields a wealth of information on decomposition kinet-
ics and mechanisms of these ingredients and propellants. To our
knowledge, a similar investigation or study of alane decompo-
sition kinetics has not been reported in the open literature.
The first objective of the present work was to develop
kinetic equations (or a model) that could accurately describe
was 45.4 ± 1.0 kJ/mol. These values indicate that alane is an
unstable compound with respect to its forming elements. Thus,
thermodynamically, alane should naturally decompose to yield
aluminium metal and hydrogen gas.
The VTS of unstabilized and stabilized aluminium hydride
◦
samples was measured at 60 C [4]. The objectives were to
develop new methods for synthesizing an “as-received” alane
then develop new methods to stabilize the product. The time
taken to reach 1% level of decomposition was ∼9.5 days for
the original sample and ∼12.7 days for the stabilized one. It is
interestingtonotethatformanyinvestigators, theVTStestswere
◦
performed at 60 C until reaching a 1% level of decomposition
by weight; however, these test conditions are somewhat arbi-
trary. In most cases, most of the work was performed using the
original “old” Taliani method [2] or a modified one. The kinetics
and mechanism of the thermal decomposition of several solvated
aluminium hydride compounds have been reported by Zakharov
and Tskhai [5]. The volume of decomposition products was
traced as a function of time at several isothermal temperatures
◦
between 50 and 100 C. For dry alane samples (that is, after the
removal of the solvents), an s-shaped type plot was obtained
during the liberation of hydrogen. The kinetics of hydrogen lib-
eration was described by a first order autocatalytic equation. The
activation energy of this step was 72.2 ± 2.5 kJ/mol.
The thermal decomposition of AlH3 and AlD3 was inves-
tigated using NMR [6]. Samples of alane were decomposed
◦
isothermally(at86–127 C);andtypicals-shapedplots, correlat-
ing percent decomposition (or amount of aluminium produced)
versus reaction time, t, were obtained. Each s-curve was divided
into three regions; the first of which corresponded to an induc-
tion period, τind, where the rate constant k1 = 1/τind. The value
of τind was defined as the time taken to achieve 5% decom-
position (or when the fraction of original alane converted to
aluminium, α, is 0.05). In the second region, an “acceleration
period” [6] was noted between α = 0.1 and 0.6; the equation
governing the kinetics in this region was α1 = k2t. The third
/3

3
4
I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
the decomposition process of alane in an inert atmosphere.
The second objective was to use the model and propose a
mechanism for alane decomposition. The third objective was
to use the model and predict the effective storage life of alane
when stored at ambient conditions.
a combination of a roughing pump and an oil diffusion pump
attached to a cryogenic liquid nitrogen trap. This system was
−7
able to pump the samples down to better than 10 Torr. The
metering valve was closed, and the entire system (reactor and
all other accessories) was suspended inside the oven. The elec-
trical wires for the components were passed through a hole at
the top of the oven and were connected to two data acquisition
modules and a power supply. The modules were connected to
feed into the serial port of a laptop computer. The experimen-
tal data were saved every 2 min; including one pressure reading
and two temperature readings as a function of time. A run was
typically continued until the sample reached the desired level of
decomposition (∼1% weight loss). These analyses often took
5–12 days.
2
. Experimental
2
.1. Materials
The as-received ␣-alane samples used in the present inves-
tigation had been synthesized and stabilized using undisclosed
technology likely similar to that described by Petrie et al. [4].
The samples had the following typical elemental analysis: aver-
age weight percentage of aluminium was 88.26%, of hydrogen
9
.96% and of carbon 0.44%. The balance, obtained by differ-
2.3. Thermal analysis (DSC and TGA experiments)
ence, 1.54% was assigned to oxygen. The traces of carbon found
by analysis, combined with a portion of the oxygen content, are
probably due to contaminated residual solvent(s) used in the
process (diethyl ether). Removal of the trapped final traces of
solvent is difficult to achieve experimentally [4].
A Mettler TGA unit (model SDTA-851e) was used with Met-
e
tler proprietary software STAR version 8. Ultra-high-purity
argon (UHP certified purity: 99.9999% Ar) was passed through
the samples at a flow rate of 60 cc/min. For the present work,
special sample containers were fabricated of quartz; the assump-
tion was made that the regular alumina sample containers used
with TGA could affect the results. Because of the differences
in thermal properties of quartz and those of the other materials,
the calibration coefficients of TGA were determined using the
quartz containers, as per the Mettler guidelines.
2
.2. Vacuum thermal stability
A new twin-chamber VTS apparatus (Fig. 1) was constructed
to investigate the thermal stability of alane. Each chamber had
the following major components: a Pyrex tube containing the
sample which was contained inside a stainless steel tube. The
tube was connected to a pressure transducer and a fine metering
valve. Two type-K thermocouples were attached on the outer
skin of the reactor. The entire system (A or B) was placed inside
Before starting a test, a sample was flushed with the selected
gas for 20 min to eliminate residual air present in the sys-
tem. The sample was then heated at a projected constant heat-
◦
ing rate, between 1 and 20 C/min from room temperature to
◦
◦
an oven regulated at a chosen temperature (normally 60 C).
350 C. After terminating a run, the density of residue was deter-
The samples were evacuated at room temperature for 24 h using
mined using a Micromeritics helium micro-Accu-Pycnometer
Fig. 1. Vacuum thermal stability dual system.

I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
35
(cell volume: 0.25 cc and expansion volume: 0.73 cc). The sur-
3.2. Surface area, morphology and particle size
face morphology of residues was examined using SEM tech-
niques.
The surface area of as-received alane, determined by krypton
adsorption, was 0.215 m /g. This value is small and suggests that
2
The surface area and porosity development of one alane
sample was studied in the following manner, using an auto-
mated surface area apparatus, Digisorb 2600, manufactured by
Micromeritics Instrument Corporation. Successive runs were
performed on one sample (starting weight = 0.85 g) while alter-
alane particles are non-porous or, at least, have an insignificant
level of porosity. Fig. 2 shows the SEM photos of the two alane
samples at two magnifications. These particles are not spherical,
as has been noted in previous published work on alane. Instead,
the particles are mostly cubes with fewer pentagonal and hexag-
onal crystal shapes. Occasionally, cracks and pits are noted on
the surface but these are not deep. The SEM size of the particles
ranges between 3 and 20 ␮m.
Assuming that all particles are cubical, the average size of
particles, and the average equivalent length of their sides, L
(in ␮m), can be estimated using the relation: L = 6/(surface
area × density). The average calculated value of L is 18.8 ␮m,
which is in reasonable agreement with SEM observations shown
in Fig. 2. It is also noted in the SEM photos that most of the par-
ticles have smooth surfaces, suggesting insignificant porosity.
◦
nating between a pre-evacuation step at 60 C for a given period
of time, and actual measurements of surface area using krypton
◦
adsorption at −196 C. The sample was always kept under vac-
uum, in He or in Kr, without ever being exposed to ambient air
for the entire set of tests. This practice insured that the generated
aluminium did not oxidize during the measurements. Details of
this test are given next.
2
.4. Calculations
Specific surface area, SA (m /g), was calculated from Kr
2
adsorption isotherms obtained at liquid nitrogen temperature
using the BET equation. The cross-sectional area of a Kr
atom adsorbed at the surface was taken as 0.210 nm /Kr atom
3.3. Vacuum thermal stability
2
[
21].
The effect of pre-evacuation time on decomposition kinetics
of alane was studied first. Several samples were pre-evacuated
at room temperature for varying periods of time: 1, 2, 10 and 16
days. There was no effect of pre-evacuation time at room temper-
For the TGA results, the data of alane decomposition were
evaluated using NETZSCH-Thermokinetics software supplied
by NETZSCH-Ger a¨ tebau GmbH in Germany.
◦
ature on subsequent decomposition kinetics at 60 C. Therefore,
a reasonable 24 h pre-evacuation time was chosen for the rest of
the work reported here. Following evacuation, the sample was
moved to the oven then heated to and held at 60 C for several
3
. Results and discussion
◦
3
.1. Characterization of as-received alane
days. The increase in reactor pressure due to hydrogen evolu-
tion, and the corresponding percent of sample weight loss were
monitored with time. When the sample reached 1% weight loss,
the experiment was normally terminated and gas analyses were
made using a mass-spectrometer. The main product gas was H2
containing small trace amounts of water.
Two selected alane samples were sent to Galbraith Labs
for elemental analysis (Al, H, and C assays). The results are
summarized in Table 1 for the dried samples. Pure alane has
a stoichiometric composition corresponding to 89.92% Al and
1
0.08% H. It is noted in the table that the two samples are short
Selected results on VTS of alane are shown in Fig. 3. For the
four runs performed, the results were reproducible and the time
taken to reach the 1% level of weight loss was in the range of
in Al and H because they contain other species such as C and
possibly oxygen. The stoichiometric Al/H ratio is 8.921 for pure
alane;theaverageexperimentalratiowas8.86. Thesampleshave
excess (organic) H that could not be accounted for. The total per-
centage of Al plus H plus C did not add up to 100%; the balance
can be assigned to other species; mostly oxygen. It was in the
range of 1.42–1.21%.
6
.8–7.6 days. An average time of 7.2 ± 0.4 days can be confi-
dently assigned to this sample.
3.4. Development of porosity during alane decomposition
Theoretically, when alane completely decomposes to yield
aluminium, the density increases from 1.486 to 2.71 g/cc.
These values were confirmed in the present investigation on
the starting alane and on the final products, using helium pyc-
nometry. From the published crystallography data, the crystal
size and structure should also be changing from hexagonal
alane (lattice parameters: a = 0.54 nm and c = 1.14 nm) [22] to
face centered cube (FCC) aluminium (a = 0.404 nm). Between
these two extremes, one would expect a combination of the
two structures, with the alane crystals breaking down and the
growth of aluminium crystals. To shed more light on this effect,
the following series of surface area measurements and porosity
evaluations were performed on alane. A fresh alane sample
Table 1
Chemical analysis and properties of alane
Chemical analysis
Sample 1
Sample 2
Al (%)
H (%)
C (%)
88.26
9.95
0.37
88.34
9.97
0.48
Total
98.58
98.79
Al/H ratio
Density (g/cc)
8.87
1.486
8.86
–
2
Surface area (m /g)
0.215
–

3
6
I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
Fig. 2. SEM photos of alane at different magnifications.
(
2
∼0.85 g) was evacuated at room temperature for 24 h; it lost
the same. The sample was again evacuated for an additional
72 h. Again, there was no weight loss or change in surface area.
This means that the 2.28% loss after the first evacuation was
due to evolution of moisture, trapped solvent (if any) and other
volatiles.
.28%. The first Kr-surface area measurement at liquid nitrogen
2
temperature was measured: 0.215 m /g. The sample was evac-
uated again at room temperature for an additional 24 h. There
was no additional weight loss, and the surface area remained
◦
Fig. 3. Vacuum thermal stability of an alane sample at 60 C.

I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
37
Fig. 4. Dependence of surface area of alane on extent of hydrogen evolution.
◦
The sample was then evacuated stepwise at 60 C for succes-
sive periods of time. After each evacuation, the sample was kept
under a blanket of ultra-high-purity He to avoid possible contam-
ination with air and to eliminate the possibility of oxidizing the
fresh nascent aluminium active sites developed during hydrogen
liberation. After each He fill up, the sample was weighed, and
a Kr-surface area measurement was executed. These two con-
secutive steps were repeated several times until the final sample
weight became constant; corresponding to the complete decom-
position of alane. That is, the remaining weight was ∼90% of
the starting weight. The results are displayed in Fig. 4. The plot
has three main regions. In the first region, between 0 and 1.9%
weight loss, the surface area is essentially the same. In this region
there is probably loss of hydrogen atoms attached to the exter-
nal surface of alane cubes plus those attached to the internal
walls of open cracks, pits and crevices. The alane particles in
this region are, most likely, intact; their sizes are essentially the
same as the starting alane. In the second region, between ∼2
and 7.5% weight loss, the particles develop porosity as a result
of the excessive hydrogen gas pressure exerted on pore walls.
The walls begin cracking, the pores begin opening and the orig-
inal crystal structure is opened. As more pores continue to open,
the surface area keeps increasing until a maximum is reached at
Fig. 5. SEM photos for (a) partially decomposed alane (3.1% weight loss) and
(b) completely decomposed alane (∼10% weight loss).
range of 12–16 ␮m, compared to 18–20 ␮m, for the starting
alane. The most interesting observation is that the original cubi-
cal and other geometrical shapes of the starting alane particles
are still preserved in the final product, cubical aluminium. The
experimental density of final residue after complete decompo-
sition (plate b) was 2.69 g/cc, which is close to the theoretical
value of 2.71 for pure aluminium. Thus, with this decomposition
reaction, it is possible to synthesize aluminium particles in the
form of cubes.
∼
7.5% weight loss. In the third region, the surface area drops
2
abruptly to ∼0.35 m /g and remains constant at this value. This
suggests that the final cubical structure of aluminium produced
may only dominate after 75% of the original hydrogen evolves.
The ratio between the final specific surface area of Al and
starting surface area of alane was approximately 1.63. This
increase is attributed to two main reasons: a decrease in particle
size after decomposition and an increase in surface roughness
during decomposition. To verify this, the surfaces of several
residues of partially and fully decomposed alane samples were
examined using SEM. Typical representative photos are shown
in Fig. 5. Plate (a) shows that the particles have developed pores
and cracks after 3.1% weight loss, and that the external surface
acquired slight roughness. Plate (b) shows the surface after the
complete decomposition of alane. The roughness has increased
drastically and there are small debris of (probably) aluminium
metal covering the surface. The average particle size, for the
larger particles in the completely decomposed sample, is the
3.5. Kinetics of alane decomposition in argon
The TGA results for alane decomposition in Ar at six heating
◦
rates (0.49, 0.98, 2.0, 4.9, 9.9 and 20.2 C/min) are shown in
◦
◦
Fig. 6. At 20.2 C/min, the decomposition began at 180 C and
was completed at 215 C, and the final weight loss was 9.9%;
◦
indicating that the entire hydrogen content of alane, bonded to
aluminium, hadbeencompletelyremoved. Forthethreerepeated
◦
runs performed at 9.9 C/min, the reproducibility of the data
was quite satisfactory, and the average weight loss was 9.05%.

3
8
I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
Fig. 6. Decomposition of alane in Ar (50 cc/min) at different heating rates.
◦
The decomposition began at a lower temperature (168 C) and
slope and intercept would yield the activation energy, Ea, and
the pre-exponential term, A. Fig. 7 displays the results obtained
with this method; the calculated values for alane decomposition
are: Ea = 97.0 ± 3.1 kJ/mol and A = 2.04 × 10 s
◦
was completed at 197 C. This trend continued with the other
heating rates. In all cases, the starting and final decomposition
temperatures were lowered, as expected, when the heating rate
decreased. The main noticeable trend is that as the heating rate
decreased, the final weight loss became smaller. That is, a sample
with a low heating rate retained more hydrogen than another one
decomposing at a higher heating rate. It appears that when alane
decomposes at a lower heating rate, the porous structure closes
earlier and a portion of the hydrogen stays trapped inside the
particles. Also, at a lower heating rate, decomposition occurs
at lower temperatures; the diffusion of hydrogen through the
particles or nucleation of aluminium sites is slow.
8
−1
.
The second MFK method is the Friedman analysis [24],
which is based on the iso-conversion method: at a given level of
decomposition, theinstantaneousdecompositionrateanditscor-
responding temperature are calculated for each heating rate. This
step is repeated at other decomposition levels until the reaction is
completed. For each decomposition level (conversion), the plot
between the logarithm of decomposition rate and the reciprocal
of decomposition temperature would normally yield a straight
line, the slope and intercept of which gives, respectively, the
values of Ea and A at this particular level of decomposition. The
computation is repeated for all other levels of decomposition and
several lines, corresponding to several iso-conversion levels, are
produced. The corresponding values of Ea and A are obtained
for each level of decomposition. The results are summarized in
Fig. 8. When the reaction involves more than one step, the con-
tribution of each step to the overall kinetics will influence the
observed activation energy. Thus, Fig. 8 simply suggests that
alane decomposition follows, at least, two steps: the first starts
and continues up to ∼0.25 conversion (∼2.5% weight loss for
alane); and the second occurs between 0.25 and 0.85 conversion
(∼2.5–8.5% weight loss for alane). This finding is in line with
the results of specific surface area reported above (as displayed
in Fig. 4).
To perform kinetic analysis in Ar atmosphere, we pursued
two independent approaches: the classical “model free kinetics”
MFK), and a more precise curve-fitting thermo-kinetics mod-
(
eling (TKM) analysis. The classical MFK approach was used
first to estimate the activation energy and the pre-exponential
factor related to the decomposition reaction. For these classical
methods, the assumption is usually made at some point of the
analysis that the decomposition is a first order reaction. Based on
the results of analysis with MFK, the more sophisticated TKM
approach was then used with as many necessary intermediate
kinetic steps and reaction orders as needed to mathematically
describe the steps involved. The model showing the lowest
possible standard deviation was assigned to the decomposition
reaction. Inthepresentinvestigation, thekineticparameterswere
estimated for different types of reactions, and finally, physical
models for the overall reaction were proposed.
The third MFK method is the Ozawa–Flynn–Wall analysis
[25,26] which also utilizes the iso-conversion approach. Simi-
lar to the Friedman analysis, a given conversion is first chosen.
For each heating rate, the corresponding temperature is noted. A
plot between the logarithm of heating rate versus reciprocal of
absolute temperature is made. A linear relationship is normally
obtained, and the corresponding values of Ea and A are computed
for this particular conversion. The computation is repeated at
other levels of conversion, and a plot is finally made between Ea
and conversion (Fig. 8). This plot for alane decomposition can
be interpreted in two different ways. First, one may argue that
3
.5.1. Model free kinetics—iso-conversion method
The classical MFK computations include the three main well-
established independent methods: the ASTM method [23], the
Friedman analysis [24], and the Ozawa–Flynn–Wall analysis
[
25,26]. With the ASTM method, an Arrhenius plot is made
between the logarithm of heating rates versus reciprocal of
absolute temperatures at the points when the maximum decom-
position rate occurs at each heating rate. If the plot is linear, the

I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
39
Fig. 7. Arrhenius plot for decomposition of alane in Ar using the ASTM method.
with this method, the activation energy of alane decomposition
is constant through the entire reaction with an average value of
In the next attempts, more complicated two-step models
were considered; the most promising included two-step consec-
utive reactions of the type: A → B → C. The two best fitting
models were chosen. They are referred to here as Model-A
and Model-B. Both models were statistically near perfection,
with correlation coefficients better than 0.998. In both cases,
Step 1 (A → B) was basically the same; it was governed by
the Kolmogorov–Johnson–Mehl–Avrami (KJMA) equation (a
modified Avrami equation), which describes mainly an nth-
dimensional nucleation or growth reaction [20,27–29]. The
KJMA model has been represented recently by the following
equation proposed by Nakamura et al. [29]:
1
03 ± 3 kJ/mol. In this case, the decomposition of alane would
follow first order kinetics. This is not the case, as will be shown
shortly. Second, as was the case with the Friedman analysis,
one may also argue that alane decomposition is more compli-
cated and involves more than one step. Both scenarios will be
considered next.
3
.5.2. Thermokinetics modeling
Additional attempts were made to fit the original raw
data to different kinetic models. The simplest model
involves first order decomposition kinetics; that is:
alane → aluminum + H2. The rate of this reaction would
be given by: (dα/dt) = A × (1 − α) × exp(−Ea/RT), where α is
the conversion fraction and R the gas constant. The analysis
indicated that alane decomposition does not follow first order
kinetics, as shown by the dotted (straight) lines in Fig. 9.
dα
(n−1)/n
= n × k(T) × (1 − α) × [−ln(1 − α)]
dt
wherek(T)istherateconstant, givenby: k(T) = A × exp(−E/RT).
Here A is the pre-exponential factor and E is its activation
energy. The Avrami exponent, n, is normally used as a tracer
for the dimensionality of the reaction. Commonly accepted
Fig. 8. Variation of activation energy with level of alane conversion to yield aluminium.

4
0
I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
Fig. 9. Comparison between the experimental data of alane decomposition (open circles), first order decomposition kinetics model (dotted lines) and two-step reaction
model (solid-line).
guidelines state that for one-dimensional growth, n ≤ 2; for two-
dimensional growth, n = 2–3; and for three-dimensional growth,
n = 3–4 [29,30]. Values of the kinetic parameters obtained here
for Step 1 in Model-A and Model-B are listed in Table 2.
The second step for Model-A and Model-B can be equally
described statistically by two different equations. For Model
A, Step 2, satisfied, once again, a second KJMA equation; the
parameters of which are displayed in Table 2. They are lower
than those of Step 1. However, for Model-B, the kinetics satisfied
KJMA equation was used to calculate the rate of each individual
reaction, dα/dt, at different temperatures and several values of
α. Fig. 10 shows the dependence of Step 1 rates on conversion
at five chosen temperatures. The rates increased with conver-
sion, reached a maximum at α = 0.45, and finally decreased
when α > 0.45. This trend is typical of a nucleation process.
The Avrami exponent, n1 is between 2 and 3 which suggests
that the first step for alane decomposition is a two-dimensional
nucleation reaction possibly taking place at the outer surface of
the particles (two-dimensional nucleation).
a typical chemical reaction (with nth order kinetics) of the type:
n
(
dα/dt) = A2 × (1 − α) × exp(−E2/RT). In this case, the order of
For Model-A, step 2 has lower values of A, E and n than
Step 1. The rates for this step are much faster than in Step 1.
Fig. 11 shows that the rates of Step 2 increased with temperature
but decreased with increasing coverage. This step is attributed
to a growth reaction; the growth of aluminium layers on the
nucleation sites already developed in Step 1. The small value
of n2, which is <1, suggests that the growth reaction is a one-
dimensional process, most likely, starting at the outer surface of
the particles and continuing inwards towards their center.
A comparison between the rates of nucleation and growth at
the reaction (power of the conversion term) was 2.52. The two
models gave excellent fitting to the raw data, as indicated by the
solid lines shown in Fig. 9. (Because the fitting is excellent, it
may be hard to see the complete solid lines at certain regions
of the plots, since the solid lines are perfectly superimposing
the raw data points.) From the statistical analysis, correlation
coefficients and standard deviation values, Model-A appeared
slightly better than Model-B. However, Model-B could not be
ruled out because its fitting to the data was also excellent.
Based on Model-A, Step 1 has a higher activation energy, pre-
exponential term and Avrami exponent than Step 2 (Table 2).
From the values of A, E, and n, obtained with each step, the
◦
100 C is illustrated in Fig. 12. This trend was valid at all other
temperatures. Since the growth rate is considerably higher than
the nucleation rate, we conclude that the nucleation (surface)
reaction is the rate-determining step that controls the overall
decomposition of alane. This suggests that if the nucleation sites
are chemically stabilized on the outer surface of the particles,
the nucleation rate will be reduced and the rate of alane decom-
position will be slowed.
Table 2
Calculated kinetic parameters for non-isothermal decomposition of alane in
argon
Model-A
Model-B
As mentioned earlier, with Model-B, Step 1 was similar to
that of Model-A; however, Step 2 was not the same. It followed
Step 1
A1 (s−1)
7
5
8
5.61 × 10
83.8
2.23
9.18 × 10
99.0
2.43
n
a simple rate law expression of the type: dα/dt ∼ (1 − α) , where
E1 (kJ/mol)
Avrami exponent: n1
n (value = 2.52) is the overall order of the second step. One pos-
sible explanation is that the aluminium atoms, once formed after
Step 1, may move over the surface and collide with each other to
form metallic bonds or to enter the crystalline structure (FCC for
aluminium). With this assumption, the aluminium atoms have to
collide with each other at a rate of two to three times, on the aver-
Step 2
A2 (s 1)
−
4.92 × 10
63.1
0.63
3.08 × 10
88.5
9
E2 (kJ/mol)
Avrami exponent: n2
Reaction order
N/A
2.52
N/A

I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
41
Fig. 10. Dependence of nucleation rates on conversion and temperature.
Fig. 11. Dependence of growth rates on conversion and temperature.
age, before they bond to each other. If this is the case, the number
of collisions at the beginning of growth process would be high
say three or four); and it will decrease, due to the proximity of
aluminium atoms, as the growth process continues. Further qual-
ification for this mechanism is beyond the scope of this article
and is left to future research.
Table 3
Prediction of weight percent of hydrogen evolved during storage of alane
(
◦
◦
◦
◦
◦
◦
Years
5 C
10 C
15 C
20 C
25 C
30 C
(a) Model-A
1
3
5
0
5
0.00
0.00
0.01
0.01
0.04
0.10
0.01
0.03
0.07
0.31
0.85
0.03
0.13
0.42
2.10
4.29
0.08
0.76
2.43
6.27
7.86
0.28
3.39
6.37
8.61
9.18
0.00
0.00
0.01
0.02
3
.6. Model prediction
1
1
With the two models developed, the prediction of alane shelf-
(
b) Model-B
◦
life during storage at 5, 10, 15, 20, 25 or 30 C was computed.
The results, which are displayed in Table 3, show two significant
trends. First, Model-A predicts longer shelf-life than Model-B.
1
3
5
0
5
0.00
0.01
0.05
0.14
0.62
1.47
0.03
0.22
0.68
2.67
4.19
0.10
1.02
2.76
5.16
6.24
0.42
3.52
5.27
7.16
8.28
1.74
5.79
7.22
9.02
9.77
0.01
0.03
0.12
0.29
1
1
◦
For example, after 3 years storage at 20 C, while Model-A pre-
dicts 0.13% weight loss of hydrogen. Model-B predicts 1.02%

4
2
I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
◦
Fig. 12. Comparison between nucleation and growth rates at 100 C.
weightloss. Itisrecalledfrom Table1thatthissamplehad9.95%
elemental hydrogen content, which is smaller than the theo-
retical value of 10.08% by 0.13%. Since this sample has been
stored in our laboratories at ambient temperature for ∼3 years,
one concludes that Model-A appears more realistic for predic-
tions than Model-B. Second, the predictions show clearly how
sensitive alane decomposition is to storage temperature fluctua-
onal alane crystals to amorphous aluminium particles (X-ray
diffraction of the aluminium left after decomposition did not
yield any peaks). Step 1 is due to a slow nucleation reaction;
possibly with the formation of aluminium nucleation sites at
the outer surface of the particles. This is the rate-determining
step for alane decomposition, and the kinetics are controlled
by an Avrami-type (KJMA) equation. The second step is
postulated as the growth of an aluminium layer towards the
center of the particles (Step 2).
◦
tions. For example, if alane is stored for 10 years at 5–10 C, the
◦
amount decomposed (as per Model-A) is insignificant; at 15 C
◦
the sample loses 0.31%, and at 30 C it loses most of its hydro-
6. Based on the mechanism of alane decomposition proposed
here, we may suggest possible methods to further slow down
the decomposition of alane. This includes coating the parti-
cles with inert materials, or formation of a protective oxide
layer at the outer surface at a proper temperature. Alane
oxidation at ambient temperature is insignificant, and the
decomposition kinetics are not affected over the material’s
storage period. However, at higher temperatures, the rate of
oxide formation is expected to be higher and may affect the
decomposition kinetics. This is the subject of another publi-
cation in this series.
gen (8.61%). Thus, the important critical element to extend the
shelf life of alane is to store it at low temperatures, preferably
◦
in the neighborhood of 10 C.
4
. Conclusions and recommendations
1
. The decomposition of alane starts with hydrogen liberation
from the external surface and pre-existing pores and cracks.
Porosity then increases, while the size of alane particles
decreases. When the decomposition is completed, the pores
are closed in the final product (aluminium).
7
. With Model-A developed, the projected decomposition of
alane in 15 years (or less) is very small, ∼0.1%, if stored
2
. The analysis of non-isothermal decomposition of alane in Ar
was performed using a popular ASTM kinetic model. This
approach is useful perhaps in yielding a reasonable single
global value of activation energy (and pre-exponential term)
that can be compared to other literature values.
◦
below 10 C. Storing alane at higher temperatures is not
recommended; the material may lose as much as its entire
◦
hydrogen content if stored for 15 years at 30 C.
8
. The conclusions reached here and the models developed are
solely based on non-isothermal decomposition of alane in an
inert atmosphere. In part II of this series (in preparation),
the isothermal decomposition of alane will be addressed
and the conclusions will be compared against those of non-
isothermal decomposition. In part III, the effect of oxidation
on alane decomposition will be addressed.
3
4
. With Freidman MFK analysis, however, alane decomposition
involves two steps; this finding is supported by the results of
surface area measurements.
. Alane decomposition can be better described by solid-state
transformation kinetics (nucleation and growth mechanism).
Two kinetic models were developed and their fit to the exper-
imental data was confirmed. The standard deviations for the
fitting were better than 0.998. The models were used to esti-
mate the shelf-life of alane when stored in inert atmospheres
for several years.
Acknowledgements
The authors want to thank Dr. Ronald Channell, Dr. Frank
Roberto, Dr. Robert Corley, Dr. Keith McFall, Dr. Donald Tzeng,
Mr. John Clark, Mr. Paul Jones, and Dr. George Harting for
5
. Alane decomposition in Ar involves two major consecutive
steps:A → B → C;representingthetransformationofhexag-

I.M.K. Ismail, T. Hawkins / Thermochimica Acta 439 (2005) 32–43
43
their fruitful discussions, suggestions and support. The techni-
[11] S. Vyazovkin, C.A. Wight, Thermochim. Acta 340–341 (1999) 53–68.
[
[
[
12] S. Vyazovkin, Int. Rev. Phys. Chem. 19 (2000) 45–60.
13] S. Vyazovkin, C.A. Wight, Chem. Mater. 11 (1999) 3386–3393.
cal assistance of Ms Marietta Fernandez (SEM examinations)
and of Ms Leslie Hudgens (density measurements) is highly
appreciated.
I.M.K. Ismail is appreciative to Late Dr. J. Opfermann (at
NETZSCH-Ger a¨ tebau GmbH) for his invaluable advice and the
effort he made in modifying the kinetics software to meet our
needs.
14] S. Vyazovkin, J.S. Clawson, C.A. Wight, Chem. Mater. 13 (2001)
60–966.
15] S. Vyazovkin, C.A. Wight, J. Phys. Chem. A 101 (1997) 8279–8284.
9
[
[16] S. Vyazovkin, C.A. Wight, J. Phys. Chem. A 101 (1997) 5653–5658.
[17] H.J. Flammersheim, J. Opfermann, Thermochim. Acta 337 (1999)
149–153.
[
[
18] J. Opfermann, W. Haedrich, Thermochim. Acta 263 (1995) 29–50.
19] B. Roduit, C. Borgeat, B. Berger, P. Folly, B. Alonso, J.N. Aebischer,
in: Proceedings of the 31st International Pyrotechnics Seminars, 11–16
July 2004, Fort Collins, Colorado, USA.
References
[
1] F.M. Brower, N.E. Matzek, P.F. Reigler, H.W. Rinn, C.B. Roberts, D.L.
Schmidt, J.A. Snovar, K. Terada, J. Am. Chem. Soc. 98 (9) (1976)
[20] M. Avrami, J. Chem. Phys. 7 (1993) 1103–1112;
M. Avrami, J. Chem. Phys. 8 (1939) 212–224;
M. Avrami, J. Chem. Phys. 9 (1940) 177–184.
[21] I.M.K. Ismail, Langmuir 8 (1992) 360.
[22] J.W. Turley, H.W. Rinn, Inorg. Chem. 8 (1969) 18.
[23] H.E. Kissinger, J. Res. Nat. Bur. Stds. 57 (1956) 217.
[24] H.L. Friedman, J. Polym. Lett. 4 (1966) 323.
[25] T. Ozawa, Bull. Chem. Soc. Jpn. 38 (1965) 1881.
[26] J. Flynn, L.A. Wall, Polym. Lett. 4 (1966) 232.
[27] A.N. Kolmogorov, Izvestiya Acad. Nauk SSSR, Ser. Math. 1 (1937)
355.
2
450–2453.
[2] M. Taliani, Gazz. Chim. Ital. 51 (1921) 184.
[3] G.C. Sinke, L.C. Walker, F.L. Oetting, D.R. Stull, J. Chem. Phys. 47
(1967) 2759.
[4] M.A. Petrie, J.C. Bottaro, R.J. Schmitt, P.E. Penwell, D.C. Bomberger,
United States Patent US 6,228,338 B1 (2001).
[
[
5] V.V. Zakharov, A.N. Tskhai, Russ. J. Inorg. Chem. 37 (9) (1992) 997.
6] V.P. Tarasov, Yu.B. Muravlve, S.I. Bakum, A.V. Novikov, Dokl. Phys.
Chem. 393 (2003) 4.
[
7] S.V. Vyazovkin, A.I. Lesnikovich, Thermochim. Acta 165 (1990)
[28] W.A. Johnson, P.A. Mehl, Trans. Am. Inst. Min. Metall. Eng. 135 (1939)
416.
2
73–280.
[
[
8] S. Vyazovkin, V. Goryachko, Thermochim. Acta 194 (1992) 221–230.
9] S. Vyazovkin, C.A. Wight, Ann. Rev. Phys. Chem. 48 (1997) 125–
[29] N. Nakamura, K. Katayama, T. Amano, J. Appl. Polym. Sci. 17 (1975)
1031.
1
49.
[30] T. Pradell, D. Crespo, N. Clavaguera, M.T. Clavaguera-Mora, J. Phys.:
Condens. Matter 10 (1998) 3833.
[
10] J. Opfermann, W. Haedrich, Thermochim. Acta 263 (1995) 29–50.