Thermal decomposition of AlH3 studied by in situ synchrotron X-ray diffraction and thermal desorption spectroscopy

Article abstract of DOI:10.1016/j.jallcom.2006.11.199

The thermal decomposition of alane was investigated by application of synchrotron X-ray diffraction (SR-XRD) and thermal desorption spectroscopy (TDS). Two polymorphs were studied, α- and γ-AlH₃. Activation energies, anisotropic volume expansions, and phase transformation paths were found. In addition, the crystal structure data, including structure of hydrogen sublattice, and small charge transfer from the aluminium towards the hydrogen sites were observed during a high-resolution SR-XRD study of α-AlH₃.

Full text of DOI:10.1016/j.jallcom.2006.11.199

Journal of Alloys and Compounds 446–447 (2007) 280–289
Thermal decomposition of AlH3 studied by in situ synchrotron
X-ray diffraction and thermal desorption spectroscopy
J.P. Maehlena, V.A. Yartysa,∗, R.V. Denysa,b, M. Fichtnerc, Ch. Frommenc,
B.M. Bulychevd, P. Pattisone, H. Emeriche, Y.E. Filinchuke, D. Chernyshove
a Institute for Energy Technology, P.O. Box 40, Kjeller No-2027, Norway
b Physico-Mechanical Institute of the National Academy of Science of Ukraine, 5, Naukova Str., 79601 Lviv, Ukraine
c Institute of Nanotechnology, Forschungszentrum Karlsruhe GmbH, P.O. Box 30640, D-76021 Karlsruhe, Germany
d Chemistry Department, Moscow State University, Leninskie Gori, Moscow 119992, MSU, Chemistry Department, Russia
e Swiss-Norwegian Beam Line, ESRF, BP 220, F-38043 Grenoble, France
Received 30 October 2006; received in revised form 28 November 2006; accepted 29 November 2006
Available online 12 January 2007
Abstract
The thermal decomposition of alane was investigated by application of synchrotron X-ray diffraction (SR-XRD) and thermal desorption spec-
troscopy (TDS). Two polymorphs were studied, ␣- and ␥-AlH3. Activation energies, anisotropic volume expansions, and phase transformation
paths were found. In addition, the crystal structure data, including structure of hydrogen sublattice, and small charge transfer from the aluminium
towards the hydrogen sites were observed during a high-resolution SR-XRD study of ␣-AlH3.
© 2006 Elsevier B.V. All rights reserved.
Keywords: Hydrogen absorbing materials; Crystal structure; Phase transitions; Synchrotron radiation
1. Introduction
ciated as an efficient way for reaching high volumetric efficiency
of the storage system and fast rates of hydrogen exchange, reach-
ing these advantages has always been paid by a price of having
a rather poor gravimetric hydrogen storage capacity.
To utilize renewable energy sources, hydrogen and electric-
ity are considered as an ideal combination of energy carriers,
providing versatile and environmentally benign solutions. Every
step in a complete hydrogen energy cycle, which includes hydro-
gen production from water by electrolysis, hydrogen storage and
conversion into electricity in the fuel cell stack, needs to be suf-
ficiently improved to increase the competitiveness of hydrogen
energy with the available energy systems. Hydrogen storage is
considered as the most important issue to be solved. An efficient
hydrogen storage solution should combine several key features
including high hydrogen storage densities, volumetric and gravi-
metric, fast and easy reversible H loading/H release, together
with affordable price and fulfilment of the safety requirements.
A storage system satisfying all these beneficial features, unfor-
tunately, is not developed yet. Even though solid state storage
materials, first of all reversible metal hydrides, have been appre-
To reach the high gravimetric hydrogen densities required
for commercially competitive hydrogen storage solutions for
the transport sector, metal hydride systems containing light
elements are needed. In the present work we have focused
our studies on the binary metal-hydrogen system of such a
type, namely on the aluminium trihydride (alane). Aluminium
trihydride alane is considered as a prospective solid H storage
material, having high gravimetric (10 wt.% H) and volumetric
density of H (2 times higher compared to LH2) and, also, because
of its convenient range of thermal stability. AlH3 forms several
polymorphic modifications, which are often polymers (AlH3)n
[1]. ␣-Alane AlH3 was first synthesized in 1947 by an exchange
reaction between AlCl3 and lithium hydride in ether solution [2].
This binary hydride has a covalent Al–H bonding and exhibits
dielectric properties. ␣-AlH3, the most stable modification of
¯
alane, has a trigonal crystal structure (space group R3c) [3,4] and
∗
is composed of the corner-sharing AlH6 octahedra. The AlH6
octahedra remain as the building elements of the crystal struc-
Corresponding author. Tel.: +47 63 80 64 53; fax: +47 63 81 29 05.
E-mail address: volodymyr.yartys@ife.no (V.A. Yartys).
0925-8388/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jallcom.2006.11.199

J.P. Maehlen et al. / Journal of Alloys and Compounds 446–447 (2007) 280–289
281
tures of the other polymorphs of AlH3. However, their packing
densities decrease compared to ␣-AlH3. This decrease is due to
the formation of cavities in ␥-AlH3 [5] or from the large empty
channels in between the octahedra in ␤-AlD3 [6]. The double
bridge Al–2 H–Al bonds were firstly experimentally observed
for the Al-based hydrides [5] in ␥-AlH3 containing the edge-
sharing AlH6 octahedral units. NMR investigation of the ␥-AlH3
[7] showed formation of distorted coordination polyhedra of H
around Al, in agreement with the data of the crystal structure
studies [5].
However, detailed mechanism of the decomposition process and
its kinetics at different temperatures requires further investi-
gation in order to propose the most optimal way of utilizing
AlH3-based systems for hydrogen supply aimed for the use in
automotive mobile applications.
This work was focused on studies of the thermal decompo-
sition of alane by application of synchrotron X-ray diffraction
(SR-XRD) and thermal desorption spectroscopy (TDS). Two
polymorphs were studied, ␣- and ␥-AlH3.
Metastable alane does not release hydrogen at normal condi-
tions and is stable in air, probably due to the formation of oxide
layers at the surface. At atmospheric pressure AlH3 desorbs
hydrogen at rather moderate temperatures (350–400 K, depend-
ing on its preparation history). Achieving sufficiently rapid and
controllable decomposition of AlH3 and proposal for efficient
synthesis routes to make the system Al–AlH3 reversible are
focused in ongoing research (see, e.g. [8–11]). Thermodynamic
characterisation of the polymorphs of alane and studies of the
kinetics of hydrogen release facilitated by alkali metal hydrides
showed that during decomposition ␤- and ␥-AlH3 transform
into the alpha modification and that significantly fast hydrogen
release from alane starts at the temperatures well below 100 ◦C.
2. Experimental details
Samples from three different alane batches were investigated in this work;
two ␣-AlH3 batches (aged ␣-AlH3 and fresh ␣-AlH3), and one ␥-AlH3. Several
methods for the wet chemical synthesis of AlH3 have been reported in the liter-
ature decades ago. The most difficult problem is to obtain AlH3 free of solvent
without decomposing the sample [1,2] as temperatures of solvent desorption
and hydride decomposition are close to each other. Both ␣-AlH3 samples were
prepared using modifications of a method described by Brower et al. [1]. Prepa-
ration method of the ␥-AlH3 modification is described in [5]. All samples were
stored under argon atmosphere, in airtight containers for the time between the
synthesis and the measurements.
Temperature desorption spectroscopy (TDS) study of the hydrogen release
from the samples were performed by heating a small amount of the selected
sample (less than 100 mg) in a stainless steel autoclave at a constant heating
Fig. 1. Thermal desorption spectra of hydrogen desorption from AlH3, (a) ␥-AlH3; (b) fresh and aged ␣-AlH3; note that the aged sample shows higher desorption
temperatures with more symmetric peak shape; probably due to thicker and more evenly distributed oxide layers; (c) integrated desorption traces for ␥-AlH3; (d)
Arrhenius plots for ␣- and ␥-AlH3. The activation energy for the ␥ is lower than for the ␣.

282
J.P. Maehlen et al. / Journal of Alloys and Compounds 446–447 (2007) 280–289
˚
Fig. 2. Rietveld-type plots of the SR-XRD pattern for ␣-AlH3 (λ = 0.37504 A) collected at room temperature showing observed (crosses), calculated (upper line)
and difference (bottom line) plots. The positions of the Bragg peaks are shown as ticks. An enlargement of the high-angle part is shown in the inset.
rate. Dynamic secondary vacuum in the measurements setup was maintained
by using a turbo-molecular pumping system, while desorption pressure was
continuously monitored by the vacuum sensor.
SR-XRD studies of AlH3 were performed at the Swiss-Norwegian Beam
Lines (SNBL) at the European Synchrotron Radiation Facility (ESRF), Greno-
ble, France. Three different experiments were performed:
part. The capillary is hermetically connected to the gas system using a carbon
ferrule mounted in a T-piece, which, in turn, is attached to the goniometer head.
Averaging over the different orientations of the crystallites, resulting in the elimi-
nationofpreferredorientationeffectsinthecollecteddiffractiondata, isachieved
by oscillating the setup around the axis of the capillary. The required elasticity of
the setup is made possible by using a flexible PEEK polymer tubing connection
between the microreaction cell and the flow system. Vacuum is created using a
turbo molecular vacuum pump. Except for the PEEK tubing allowing the oscil-
lation of the sample cell, stainless steel tubes are used for the connections to
prevent oxygen diffusion through the tubes during the experiments.
The diffractometer at station BM01B (Si(1 1 1) channel-cut monochromator,
scintillation detectors) is equipped with six counting chains, with an angular
offset in 2θ of ∼1.1◦. For in situ measurements, in order to keep the counting
time per scan as low as possible, the detector bank is moved by 1.2◦ during
one measurement and the data from the six different detectors are added using
a data-binning program. One data set (2θ = 6.07–12.67◦) was collected in 2 min
during the heating of the sample under secondary vacuum conditions. The high-
resolution measurement of the aged ␣-AlH3 was performed in steps of 0.005◦
1. High-resolution scan atroomtemperatureof theaged ␣-AlH3 sample (station
BM01B).
2. In situ scan of the aged ␣-AlH3 sample (station BM01B), heating rate
1 K/min.
3. In situ scan of the ␥-AlH3 sample (station BM01A), heating rate 1/2 K/min.
The in situ measurements were performed using a setup designed for in situ
studies of the chemical processes occurring in hydrogen or gas/vacuum atmo-
sphere [12–14]. In this setup, a small amount of the sample is put in a 0.7 mm
quartz glass capillary filling approximately 1–2 mm of the capillary’s bottom
Fig. 4. In situ SR-XRD pattern of the thermal decomposition of ␣-AlH3 (heating
rate 1 ◦C/min). The ␣-AlH3 peaks are diminishing during the heating, indicating
amorphisation. At higher temperatures small aluminium peaks appear indicating
a nucleation and growth process for the aluminium formation.
Fig. 3. Fourier transform of the observed SR-XRD pattern (GSAS Rietveld
refinements) showing the Al–H bonds (equatorial plane of the octahedron con-
taining 4 H atoms is selected).

J.P. Maehlen et al. / Journal of Alloys and Compounds 446–447 (2007) 280–289
283
◦
◦
˚
including the H position gave the following best goodness-of-fit
parameters: Rwp = 14.2%, Rp = 11.7%, and χ2 = 6.0. Including
the hydrogen atom (at site 18e [0.625(2), 0, 1/4]), the quality
of the fit significantly improved, giving goodness-of-fit parame-
ters of Rwp = 12.1%, Rp = 10.6%, and χ2 = 4.9. Fig. 2 shows the
Rietveld-type plots of the high-resolution data.
in the 2θ range from 5.540 to 36.935 . A wavelength of 0.37504(2) A was used
in both experiments.
Experimental data at BM01A were collected using a MAR2300 image
plate detector and then integrated using the software Fit2D. The wavelength
(λ = 0.71118 A) and sample to detector distance (d = 199.768 mm, giving a
usable 2θ range of about 1–40◦ for the 1D patterns) were calibrated from an
˚
individual run of a LaB6.
All powder diffraction data were analysed by the Rietveld whole-profile
refinement method [15] using the General Structure Analysis System (GSAS)
software [16]. Peak shapes were described by a multi-term Simpson’s rule inte-
gration of the pseudo-Voigt function [17,18], which includes the asymmetry
correction according to Finger et al. [19]. Anisotropic line broadening of the
diffraction pattern was observed for the ␥-polymorph and was accounted for by
refinements using the Thompson–Cox–Hastings pseudo-Voigt type function as
proposed in [20].
The refinements yielded an Al–H bonding distance of
1.712(3) A and an octahedral coordination of Al into AlH6 units
˚
where bridge bonds Al–H–Al are formed with a bond angle of
141◦. Furthermore, they indicated a small charge transfer from
Al to H corresponding to the formation of Al+0.15 and H−0.05
.
This charge distribution can be observed in the Fourier transform
of the observed XRD pattern as shown in Fig. 3. We note a good
3. Results and discussion
3.1. Thermal desorption spectroscopy
Fresh ␣-, and ␥-AlH3 samples were studied by TDS at three
different constant heating rates (ꢀT = 0.5, 1, and 2 K/min). A
two-peak desorption behaviour was observed for all runs (see
Fig. 1a and b). Prior to the main desorption peak which appeared
at around 130–160 ◦C, a smaller, low temperature desorption
event was observed. It is well pronounced as a separate des-
orption peak in case ␥-AlH3 (Fig. 1a) and seen as a shoulder
on a low-temperature side of the peak observed for the ␣-AlH3
(Fig. 1b). For ␥-AlH3 it is reasonable to suggest that the low-
temperature event corresponds to a direct decomposition process
␥-AlH3 → Al + 3/2 H2; the second, main event, is associated
with a decomposition of ␣-AlH3, which is formed first during
the heating of ␥-AlH3. In case of ␣-AlH3, since hydrogen release
from this modification should be a single step process, it is likely
that the observed behaviour is due to non-uniform surface prop-
ertiesofthedifferentpartsofthesamplebecauseofthevariations
in the covering by the particles protective oxide layer.
Activation energies for the desorption process from AlH3
were found from Arrhenius plots: Ea = 136 and 92 kJ/mol AlH3,
for ␣-AlH3 and ␥-AlH3, respectively. Graetz and Reilly reported
that the activation energy might vary with the age of the sample.
They found between 102 and 150 kJ/mol for the ␣-phase [10].
A TDS run of the aged ␣-AlH3 sample was also performed.
Compared to fresh samples of ␣-AlH3, a much higher desorption
temperature was observed for the aged sample. Experimental
data and the Arrhenius plots of the TDS experiments are pre-
sented in Fig. 1.
3.2. High-resolution SR-XRD investigation of α-AlH3
The high-resolution SR-XRD pattern of ␣-AlH3 was indexed
¯
in trigonal setting, space group R3c (No. 167). The refinements
˚
˚
yieldedunitcellparameters:a = 4.44994(5) A;c = 11.8200(2) A,
3
˚
and V = 202.701(4) A which agree well with the reference data:
Fig. 5. Results from the refinements of the in situ SR-XRD patterns of the aged
␣-AlH3 thermal decomposition vs. temperature; (a) volumes of the unit cells
of the crystal structures of alane and Al metal (per one mole AlH3/Al), (b)
evolution of the unit cell parameters of the hydride during the heating, and (c)
temperature-dependent fraction of the crystalline aluminium hydride (by mass)
in the system (the refinements included crystalline aluminium and crystalline
aluminium hydride).
3
˚
˚
˚
a = 4.449 A; c = 11.804 A; V = 202.34 A [3]. Since Al has a
relatively low atomic number, it was possible to locate the
hydrogen sublattice during the refinements, starting with Al
placed in a special position 6a [0, 0, 0] and performing dif-
ference Fourier analysis. Performing the refinements without

284
J.P. Maehlen et al. / Journal of Alloys and Compounds 446–447 (2007) 280–289
correspondence between the data of the present powder XRD
work and the results of the powder neutron diffraction study of
AlD3 [3], where an Al–D bond distance of 1.715 A in the AlD6
(see Fig. 4). Rietveld-type refinements of the data were per-
formed (goodness-of-fit parameters for the refinements were
all rather large, with Rwp in the range 13–28%) and showed
that on heating a linear increase of the unit cell dimensions of
AlH3 takes place in an interval from RT to ∼125 ◦C (a = 4.46;
˚
octahedra was reported.
Hydrogenation is accompanied by a volume increase of
3
◦
˚
˚
103.5 (!)% equivalent to 5.73 A /atom H. We note that these
numbers are significantly higher compared to the conven-
tional metal/intermetallic hydrides where both corresponding
values are significantly lower, respectively, 20–30% and
c = 11.83 AforT = 135 C;seeFig. 5aandb). Thevolumeexpan-
sion is pronouncedly anisotropic, as can be seen from Fig. 5b.
Indeed, the c-axis is almost constant in the temperature inter-
val 25–125 ◦C (ꢀc/c ∼ −0.02%). This contrasts to a significant
expansion along [1 0 0]: ꢀa/a ∼ 0.28%. The latter expansion
can mainly be attributed to the elongations of the bridge bonds
3
˚
2.5–3.0 A /atom H [21,22]. This dramatic increase of the unit
cell volume is accompanied with an enlargement of the closest
¯
˚
Al–Al distances of 13.3% (growing from 2.86 A in the alloy to
Al–H–Al aligned along [1 1 1/2], which are expanding from
˚
˚
3.24 A in the hydride).
3.243 to 3.248 A (ꢀd/d ∼ 0.17%).
During the heating, a continuous decrease of the intensi-
ties of the pattern of AlH3 was observed indicating a gradual
decomposition of the crystalline hydride phase, probably form-
ing amorphous aluminium. This decrease was not accompanied
by any significant broadening of the peaks. The intensities of the
diffraction peaks from Al firstly appeared at ∼135 ◦C and then
3.3. In situ SR-XRD investigation of the α-AlH3 to Al
transformation
The in situ diffraction pattern of the decomposition of the
hydride were collected in the temperature range 25–150 ◦C
Table 1
Selected parameters derived from Rietveld-type refinements of the in situ SR-XRD data of the thermal decomposition of AlH3
Parameters
In situ SR-XRD study of thermal decomposition of ‘aged’ ␣-AlH3
T (◦C)a
30
80
100
4.4636(3)
115
4.4672(2)
127.5
4.4681(4)
137.5
4.468(2)
b
␣-AlH3
˚
a (A)
4.4581(1)
4.4614(2)
˚
c (A)
11.8420(7)
203.82(1)
33.970(2)
100
11.8497(9)
204.09(1)
34.015(2)
100
11.838(1)
204.26(2)
34.043(3)
100
11.841(1)
204.64(1)
34.107(2)
100
11.839(2)
204.64(3)
34.113(4)
100
11.82(1)
204.4(1)
34.06(2)
42.6
3
˚
V (A )
3
c
˚
ν (A )
wt.%
No. var.d
14
13
10
10
10
13
Rwp (%)
17.68
16.61
18.17
18.87
17.22
16.26
In situ SR-XRD study of thermal decomposition of ␥-AlH3
T (◦C)a
30
80
88
95
100
115
b
␣-AlH3
˚
a (A)
4.4438(3)
4.4493(3)
4.4524(1)
11.8110(6)
202.77(1)
33.796(2)
32.49
4.4530(1)
11.8111(4)
202.829(7)
33.805(1)
44.00
4.4534(1)
11.8110(5)
202.858(8)
33.810(1)
38.91
4.4525(6)
˚
c (A)
11.807(1)
201.92(3)
33.653(4)
15.97
11.810(1)
202.48(2)
33.746(4)
16.354
11.809(3)
202.75(5)
33.791(8)
6.65
3
˚
V (A )
3
c
˚
ν (A )
wt.%
e
␥-AlH3
˚
a (A)
5.3762(4)
7.3431(6)
5.7684(3)
227.72(2)
37.954(4)
70.62
5.3757(6)
7.3447(7)
5.7708(4)
227.85(3)
37.975(4)
68.83
5.3758(5)
7.3450(8)
5.7711(4)
227.87(3)
37.979(5)
50.56
5.376(1)
7.344(2)
5.7707(8)
227.82(6)
37.97(1)
19.50
5.375(4)
7.343(7)
5.771(4)
227.8(2)
37.97(4)
3.32
–
–
–
–
–
< 0.5
–
–
˚
b (A)
˚
c (A)
3
˚
V (A )
3
c
˚
ν (A )
wt.%
x2
0.7886(5)
0.0830(4)
0.7882(6)
0.0831(5)
0.7873(7)
0.0829(6)
0.785(1)
0.082(1)
0.725(8)
0.087(6)
y2
No. var.d
37
37
37
37
34
26
Rwp (%)
4.26
4.25
3.79
3.76
3.82
4.4
Standard deviations are given in parentheses.
a
Estimated error in temperature is ∼ 5 ◦C.
b
¯
Crystal structure data for ␣-AlH3. Space group R3c; 6 Al in 6a (0, 0, 0); 18 H in 18e (0.63, 0, 0).
Molar volume (unit cell volume/6 f.u. AlH3).
c
d
e
No. of variables refined in the Rietveld-type refinements.
Crystal structure data for ␥-AlH3 [5]. Space group Pnnm; 2 Al1 in 2b (0, 0, 1/2); 4 Al2 in 4g (x2, y2, 0); 2 H1 in 2d (0, 1/2, 1/2); 4 H2 in 4g (0.16, 0.24, 1/2); 4
H3 in 4g (0.43, 0.64, 1/2); 8 H4 in 8h (0.79, 0.08, 0.30).

J.P. Maehlen et al. / Journal of Alloys and Compounds 446–447 (2007) 280–289
285
Fig. 6. In situ SR-XRD pattern of ␥-AlH3 thermal decomposition (in the temper-
ature range 50–100 ◦C, heating rate 1/2 ◦C/min). The ␥-AlH3 peaks observed
in the low temperature region diminish as ␣-AlH3 starts to form. At higher
temperatures, a relatively fast growth of the aluminium diffraction peaks is
observed.
Fig. 7. Rietveld-type plots of the ␥-AlH3 sample from the in situ scan
◦
˚
(λ = 0.71118 A) collected at 92 C showing observed (crosses), calculated (upper
line) and difference (bottom line) plots. The positions of the Bragg peaks are
shown as ticks (from top to bottom: ␥-AlH3, ␣-AlH3, Al).
gradually increased with raising temperature. At T ∼ 145 ◦C,
the hydride peaks completely vanished indicating completeness
of the transformation AlH3 → Al. The fraction of the hydride
observed in the SR-XRD patterns as a function of the temper-
ature is given in Fig. 5c. These in situ data agree well with the
TDS measurements showing the onset of the peak of H desorp-
tion at ∼140 ◦C as can be seen in Fig. 1b. Crystallographic and
Rietveld refinement parameters for selected scans are presented
in Table 1.
aluminium. Goodness-of-fit parameters for the refinements
were all satisfactory, with Rwp in the range 3.7–4.7%. A plot
of a selected range of the experimental data is presented in
Fig. 6. Crystallographic and Rietveld refinement parameters
for selected scans are presented in Table 1 and an example of
Rietveld-type plot (data collected at T = 92 ◦C) is presented in
Fig. 7, where similar abundances of ␥- (34%) and ␣- (42%)
are observed, with 17% of the initial alane already decomposed
into Al.
3.4. In situ SR-XRD investigation of the thermal
decomposition of γ-AlH3
Fig. 8 presents the obtained phase fractions and calculated
hydrogen concentrations as a function of temperature while
Figs. 9 and 10 show the evolution of the unit cell volumes
and unit cell parameters for the two alane polymorphs as a
function of the temperature. The ␥-AlH3 starts to decompose
The initial sample used in the in situ SR-XRD experiment
contained mainly ␥-AlH3 with minor additions of ␣-AlH3 and
Fig. 8. Calculated phase percentages and hydrogen concentration from the in situ SR-XRD patterns of the ␥-AlH3 thermal decomposition.

286
J.P. Maehlen et al. / Journal of Alloys and Compounds 446–447 (2007) 280–289
Fig. 9. Changes in unit cell dimensions for the ␣-AlH3 phase vs. temperature from the in situ SR-XRD pattern of the ␥-AlH3 thermal decomposition. The numbers
indicates the partition into temperature intervals with constant lattice changes: interval 1: (25–80 ◦C) constant phase fractions, mainly ␥-AlH3 present, 2: (80–88 ◦C)
fast growth of ␣-AlH3, growth of aluminium, ␥-AlH3 decreases, 3: (88–95 ◦C) slower growth of ␣-AlH3, ␥-AlH3 decreases, fast growth of aluminium, 4: (95–115 ◦C)
␣-AlH3 decreases to ∼0 at 115 ◦C, ␥-AlH3 decreases to ∼0 at 100 ◦C, steady growth of aluminium.
at T ∼ 80 ◦C and the content of aluminium and ␣-AlH3 starts
to increase. As this happens, the anisotropic expansion of the
␣-AlH3 unit cell gets more pronounced, as can be seen from
Fig. 9. The unit cell axes expand almost linearly within lim-
ited temperature regions. Relative changes per degree Celsius
(ꢀF/ꢀT) were obtained by linear fits of the values obtained
from the Rietveld-type refinements within four chosen temper-
ature partitions and are presented in Table 2 and graphically for
the ␣-modification in Fig. 11. In the first partition (25–80 ◦C)
before the decomposition of ␥-AlH3 sets in, a change ꢀF/ꢀT of
fast growth-regime of ␣-AlH3, ꢀF/ꢀT for the a-axis increases
almost four-fold to 0.89·10−2 ◦
C
−1. The maximum formation
rate of ␣-AlH3 appears at 88 ◦C, and after this, in partition 3
(88–95 ◦C), the axes expansion slows down to values similar as
for the first partition. At higher temperatures (95–115 ◦C), parti-
tion 4, both the amount of ␣-AlH3 and ␥-AlH3 decreases to ∼0
(at ∼115 and ∼100 ◦C, respectively) and a possible small nearly
isotropic contraction of the ␣-AlH3 unit cell is observed. Rep-
resentative Rietveld-plots for the different temperature regions
are presented in Figs. 7 and 12.
0.23 × 10−2 ◦
C
−1 is observed for the a-axis of ␣-AlH3 while the
To get a more detailed view of how the phase transfor-
mations proceed, the changes of hydrogen concentrations in
c-axis is almost constant. In partition 2 (80–88 ◦C), during the
Fig. 10. Changes in unit cell dimensions for the ␥-AlH3 phase vs. temperature from the in situ SR-XRD patterns of the ␥-AlH3 thermal decomposition.

J.P. Maehlen et al. / Journal of Alloys and Compounds 446–447 (2007) 280–289
287
Table 2
Changes in lattice parameters as obtained from Rietveld-type fittings of in situ SR-XRD data of the thermal decomposition of AlH3 samples
a
a
Phase
Variable (unit)
Tinterval (◦C)
Xstart
Xfinal
ꢀF/ꢀTb (10−2%/◦C)
Notes
c
˚
␣-AlH3
a (A)
25–127.5
4.4567
4.4671
0.23
Only ␣-AlH3 present AlH3 (␣-decomposition
starts at ∼128 ◦C)
˚
c (A)
11.8405
203.67
11.8392
204.60
−0.01
3
˚
V (A )
0.44
d
˚
␣-AlH3
a (A)
30–80
4.4438
4.4490
0.23
Mainly ␥-phase present (constant phase
fractions)
˚
c (A)
11.8074
201.93
4.4493
11.8103
202.47
4.4525
11.8105
202.46
4.4525
11.8108
202.77
4.4531
0.05
0.52
0.89
0.06
1.85
0.18
3
˚
V (A )
˚
a (A)
80–88
88–95
␥-Phase starts to decompose, ␣-phase grows fast
˚
c (A)
3
˚
V (A )
˚
a (A)
Slower ␣-phase growth (␣ formation peak
∼88 ◦C)
˚
c (A)
11.8112
202.79
4.4534
11.8113
202.84
4.4529
0.01
0.38
−0.06
3
˚
V (A )
˚
a (A)
95–115
30–75
␥-Phase < 1 wt.% above 105 ◦C. ␣-phase
decreases
˚
c (A)
11.8114
202.87
11.8095
202.79
−0.08
−0.20
−0.02
3
˚
V (A )
d
˚
␥-AlH3
a (A)
5.37622
5.37579
Mainly ␥-phase present (constant phase
fractions)
˚
b (A)
7.34321
5.76853
227.73
7.34474
5.77069
227.85
0.05
0.08
0.11
˚
c (A)
3
˚
V (A )
a
Value of parameter obtained from linear fitting of lattice parameters in the range given in column ‘Tinterval’. Xstart is the first value in the temperature interval,
Xfinal, the last. The lattice parameters used as data points in the fitting were obtained from Rietveld-type refinements of the experimental diffraction data.
b
The fractional changes ꢀF/ꢀT are defined as 100·ꢀX/T where ꢀX is 100·(Xfinal − Xstart)/(Xstart).
c
As obtained from the in situ SR-XRD study of thermal decomposition of ‘aged’ ␣-AlH3.
d
As obtained from the in situ SR-XRD study of thermal decomposition of ␥-AlH3.
Fig. 11. Relative changes in unit cell dimensions for the ␣-AlH3 phase vs. temperature from the in situ SR-XRD patterns of the ␥-AlH3 thermal decomposition.

288
J.P. Maehlen et al. / Journal of Alloys and Compounds 446–447 (2007) 280–289
Fig. 12. Selected Rietveld-type plots of the in situ SR-XRD pattern of the ␥-AlH3 thermal decomposition showing observed (crosses), calculated (upper line) and
difference (bottom line) plots. The positions of the Bragg peaks are shown as ticks (from top to bottom: ␥-AlH3, ␣-AlH3, Al, impurity phases).
the hydride phases were calculated based on the weight frac-
tions of the crystalline phases as obtained from the Rietveld
refinements and assuming that hydrogen content decreases from
10 wt.% H to 0 during the decomposition process. The results
are presented in Fig. 13 in the form of ‘hydrogen desorption
traces’. Formation of the ␣-phase consumes less hydrogen than
the number of moles of hydrogen liberated by the decomposi-
tion of ␥-AlH3. Obviously, an excess of hydrogen is desorbed
directly from the ␥-AlH3, i.e. due to direct transformation of
␥-AlH3 to aluminium. This provides an evidence of coexistence
of two decomposition paths: ␥-AlH3 → Al, and ␥-AlH3 → ␣-
AlH3 → aluminium. From integrating the ‘desorption traces’,
we arrive to a conclusion that direct decomposition of ␥-AlH3 is
a prevailing process, yielding about 60% of the released hydro-
gen at the given heating rate (1/2 ◦C/min).
4. Summary
TDS studies and subsequent Arrhenius plots of fresh ␣-AlH3
and ␥-AlH3 revealed that the activation energy of hydrogen des-
orption from ␥-AlH3, 92 kJ/mol AlH3, is lower compared to
␣-AlH3, 136 kJ/mol AlH3. A TDS run of the aged ␣-AlH3 sam-
ple showed a much higher desorption temperature compared to
the freshly synthesized sample.
Crystal structure data for ␣-AlH3, including the struc-
ture of hydrogen sublattice, was successfully obtained by
Rietveld-type refinements and difference Fourier analysis of
the high-resolution SR-XRD data. The refinements indicated
a small charge transfer from Al to H atoms corresponding to the
formation of Al+0.15 and H−0.05
.
A significant anisotropic volume expansion of ␣-AlH3 dur-
ing its heating was observed with the main expansion proceeding
along the a-axis and attributed to the elongations of the bridge
bonds Al–H–Al aligned along [1 −1 1/2]. The expansion of the
a-axis proceeds considerably faster during the fast “in situ” for-
mation of the ␣-phase from the ␥-hydride, as observed in the
decomposition experiments of the ␥-AlH3. A small anisotropic
expansion of the ␥-AlH3 was also observed.
Fig. 13. Calculated hydrogen desorption and absorption traces from the in situ
SR-XRD patterns of the ␥-AlH3 thermal decomposition. Subtracting the amount
of hydrogen used during the formation of ␣-AlH3 from the amount of hydrogen
released by the decomposition of ␥-AlH3, a net hydrogen desorption trace is
found (shaded area). This amount is attributed to direct ␥-AlH3 → Al phase
transformation.

J.P. Maehlen et al. / Journal of Alloys and Compounds 446–447 (2007) 280–289
[3] J.W. Turley, H.W. Rinn, Inorg. Chem. 8 (1969) 18.
289
In the decomposition experiment of the ‘aged’ ␣-AlH3 sam-
ple, it seems that formation of an amorphous phase precedes the
formation of aluminium and subsequently that the nucleation
step limits the rate of aluminium formation. Aluminium forms
easily in the ␥-AlH3 decomposition experiment. However, in
this sample aluminium is already present of in the starting
composition, providing already existing nuclei for the growth
of aluminium. Thus, both experiments give complementary
evidences that nucleation is the rate-limiting step in the growth
of aluminium during the alane decomposition process. This
observation agrees well with the findings of Graetz and Reilly
[10].
FromtheanalysisoftheinsituSR-XRDpatternofthethermal
decomposition of ␥-AlH3, evidence of two parallel decomposi-
tion processes of ␥-hydride to aluminium were found (1) ␥ → Al
and (2) ␥ → ␣ → Al. Direct decomposition process ␥ → Al was
identified as a prevailing one.
[4] J.P. Maehlen, V.A. Yartys, in: D. Chandra, J. Petrovic, R.G. Bautista, A.
Imam (Eds.), Symposium TMS 2006 Annual Meeting on Advanced Mate-
rials for Energy Conversion III, San Antonio, Texas, USA, March 12–16,
2006, pp. 77–85.
[5] V.A. Yartys, R.V. Denys, J.P. Maehlen, C. Frommen, M. Fichtner, B.M.
Bulychev, H. Emerich, Inorg. Chem. 46 (2007) 1051–1055.
[6] H.W. Brinks, W. Langley, C.M. Jensen, J. Graetz, J.J. Reilly, B.C. Hauback,
J. Alloys Compd., in press.
[7] S.-J. Hwang, R.C. Bowman, J. Graetz, J.J. Reilly, Mater. Res. Soc. Symp.
Proc. 927 (2006), 0927-EE03-03 and S.-J. Hwang, R.C. Bowman, Jr., J.
Graetz, J.J. Reilly, MH2006, O-071.
[8] G. Sandrock, J. Reilly, J. Graetz, W.M. Zhou, J. Johnson, J. Wegrzyn, Appl.
Phys. A: Mater. Sci. Process. 80 (2005) 687–690.
[9] J. Graetz, J.J. Reilly, J. Alloys Compd. 424 (1/2) (2006) 262–265.
[10] J. Graetz, J.J. Reilly, J. Phys. Chem. B 109 (2005) 22181–22185.
[11] G. Sandrock, J. Reilly, J. Graetz, W.-M. Zhou, J. Johnson, J. Wegrzyn, J.
Alloys Compd. 421 (1/2) (2006) 185–189.
[12] M. Stange, J.P. Maehlen, V.A. Yartys, P. Norby, W. van Beek, H. Emerich,
J. Alloys Compd. 404 (2005) 604–608.
[13] P. Norby, J. Am. Chem. Soc. 119 (1997) 5215–5221.
[14] E.K. Andersen, I.G.K. Andersen, P. Norby, J.C. Hanson, J. Solid State
Chem. 141 (1998) 235–240.
Acknowledgements
[15] H.M. Rietveld, J. Appl. Crystallogr. 2 (1969) 65–71.
[16] A.C. Larson, R.B.v. Dreele (LANL, Los Alamos, 1994).
[17] P. Thompson, D.E. Cox, J.B. Hastings, J. Appl. Crystallogr. 20 (1987)
79–83.
[18] C.J. Howard, J. Appl. Crystallogr. 15 (1982) 615–620.
[19] L.W. Finger, D.E. Cox, A.P. Jephcoat, J. Appl. Crystallogr. 27 (1994)
892–900.
This work received support from the Norwegian Research
Council. We are grateful to Dr. M. Sato (Tokai University,
Japan and IFE), T. Foerde (IFE) and Dr. A.B. Riabov (PhMI
AS Ukraine) for their help.
References
[20] P.W. Stephens, J. Appl. Crystallogr. 32 (1999) 281–289.
[21] V.A. Yartys, V.V. Burnasheva, K.N. Semenenko, Usp. Khim. 52 (1983)
529–562.
[22] V.A. Yartys, V.V. Burnasheva, K.N. Semenenko, N.V. Fadeeva, S.P.
Solovev, Int. J. Hydrogen Energy 7 (1982) 957–965.
[1] F.M. Brower, N.E. Matzek, P.F. Reigler, H.W. Rinn, C.B. Roberts, D.L.
Schmidt, J.A. Snover, K. Terada, J. Am. Chem. Soc. 98 (1976) 2450–2453.
[2] A.E. Finholt, A.C. Bond, H.I. Schlesinger, J. Am. Chem. Soc. 69 (1947)
1199–1203.