Nanocrystalline alanates - Phase transformations, and catalysts

Article abstract of DOI:10.1016/j.jallcom.2004.12.178

X-ray absorption in combination with powder X-ray diffraction and surface analytical studies were performed in order to investigate the chemical state, local order and spatial distribution of Ti in sodium alanate doped with TiCl₃ and small Ti clusters. Kinetic investigations of the different hydrogenation and dehydrogenation steps of doped NaAlH₄ indicate a nucleation and growth mechanism as rate limiting step in the transformation of the material. The results support a mechanism where Ti has an accelerating effect on the phase transformation, probably by introducing non-thermal defects into the material.

Full text of DOI:10.1016/j.jallcom.2004.12.178

Journal of Alloys and Compounds 404–406 (2005) 732–737
Nanocrystalline alanates—Phase transformations, and catalysts
∗
´
M. Fichtner , P. Canton, O. Kircher, A. Leon
Karlsruhe Research Centre, Institute of Nanotechnology, PO Box 3640, D-76021 Karlsruhe, Germany
Received 7 September 2004; accepted 2 December 2004
Available online 15 July 2005
Abstract
X-ray absorption in combination with powder X-ray diffraction and surface analytical studies were performed in order to investigate the
chemical state, local order and spatial distribution of Ti in sodium alanate doped with TiCl3 and small Ti clusters. Kinetic investigations of
the different hydrogenation and dehydrogenation steps of doped NaAlH4 indicate a nucleation and growth mechanism as rate limiting step in
the transformation of the material. The results support a mechanism where Ti has an accelerating effect on the phase transformation, probably
by introducing non-thermal defects into the material.
© 2005 Elsevier B.V. All rights reserved.
Keywords: Metal hydrides; Hydrogen absorbing materials; Aluminum hydrides; X-ray diffraction
1. Introduction
time, sodium alanate has evolved into a model system for the
research community and efforts are being made to finding
out more about the details of the transformation mechanism
of functional nanocomposites based on NaAlH4 as hydrogen
carrier. In particular, possible intermediates in the transfor-
mation, the rate-determining reaction steps, and the role of
Ti dopants are subject of research.
The present work aims at giving an insight into the fate
of Ti precursors during ball milling and dehydrogenation cy-
cling. Moreover, a detailed analysis of XRD patterns and
depth profiling of doped alanate sample are performed to ob-
tain information about the spatial distribution of Ti and effects
of ball milling. Finally, the relationship between the kinetic
measurements made and the state and possible role of Ti in
the material will be highlighted.
After the first stimulating work of Bogdanovic et al. [1],
researchers have investigated reversible hydrogen storage
properties of sodium alanates (e.g. [2–5]). A succeeding
phenomenological study showed that ball milling of NaAlH4
has a remarkable effect in speeding up the decomposition
kinetics of the pure compound [2]. Moreover, improvements
were reached when the alanate was mixed with a dopant
and subjected to ball milling rather than wet impregnation
proposed in [1]. Nanocomposites prepared by ball-milling
NaAlH4 with Ti compounds such as TiCl3 or Ti alkoxides
as catalyst precursor showed superior hydrogenation and
dehydrogenation behavior [4]. The latest improvements
have been achieved by using a small Ti cluster as dopant
[7,8]. The addition of 5 mol% Ti on the basis of the cluster
allows a rehydrogenation to 80% of maximum value within
50 s only at 100 ◦C and 100 bar hydrogen pressure [9].
While there has been a remarkable progress in the devel-
opment of dopants, only a few attempts of developing suit-
able alanate-based H carrier materials with higher reversible
hydrogen contents have been reported [6,10,11]. In the mean-
2. Experimental
All sample preparations were done in an argon-filled
glove box equipped with a recirculation system to keep the
water and oxygen concentrations below 1 ppm. Chemical
operations were performed on the bench under a nitrogen
atmosphere using Schlenk tube techniques. Ball milling was
performed with a Fritsch P6 planetary mill at 600 rpm using
∗
Corresponding author. Tel.: +49 7247 82 5340; fax: +49 7247 82 6368.
E-mail address: fichtner@int.fzk.de (M. Fichtner).
0925-8388/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.jallcom.2004.12.178

M. Fichtner et al. / Journal of Alloys and Compounds 404–406 (2005) 732–737
733
silicon nitride balls and a vial with a ball to powder weight
ratio of about 20:1. The vial was filled and sealed in the glove
box. TiCl3 (Sigma–Aldrich, 99.999% purity) was used as
received. Purification of NaAlH4 and the synthesis of Ti clus-
ters are described in detail in Ref. [7]. Experimental details
concerning the X-ray absorption and X-ray photoelectron
measurements can be found elsewhere [12,19].
For XRD experiments, the powder was prepared and mea-
sured under an argon atmosphere in a special XRD sample
holder. A Philips X’PERT diffractometer equipped with a fo-
cusing graphite monochromator on the diffracted beam and
a proportional counter was used. Moreover, a divergence and
antiscatter slit of 0.5◦, a receiving slit of 0.2 mm, and Ni-
filtered Cu K␣ radiation (40 mA, 50 kV) were employed. The
patterns were recorded at 295 K with a step size of 0.05◦ in a
10–140◦ 2θ range using a counting time of 5 s per run. Total
counting time per step was 30 s.
Powder samples were prepared for SIMS and SNMS mea-
surements by pressing small amounts of doped alanate pow-
der in In foil (99.999%) and mounting a 1 cm2 piece under a
Ta mask (orifice 7 mm) on a sample holder of an INA-3 instru-
ment by SPECS GmbH. The samples were transferred to the
instrument under an Ar atmosphere, and the mounted sample
holder was cooled with a liquid nitrogen flow for the mea-
surements. SNMS experiments were conducted as described
in [17]. Charging the non-conductive samples was avoided
by performing the measurements in a high-frequency sputter
mode. At target potential time, Ar+ ions are attracted from
the Ar plasma in front of the sample to sputter the sample. At
zero potential time, plasma electrons are attracted from the
plasma to compensate charged sample areas.
Fig. 1. Normalized Ti K-edge XANES spectra of NaAlH4 doped with
5 mol% of Ti (Ti13 · 6THF basis) at different stages. Spectra of pure Ti metal
and Ti13 · 6THF are shown for comparison.
The presence of the beat node at high k-space range as well
as the fail of the refinement with Al as backscatterers around
the Ti atoms suggests that the formation of an alloy between
Ti and Al in the first cycle is not supported. The comparison
of those data with references such as TiAl, Ti3Al or TiAl3
will be published in a forthcoming paper.
3.1.2. Ti13· 6THF doped samples
The Ti13 · 6THF structure consists of a metal core having
an icosahedral shape and six THF ligands. The O atom of
each THF molecule is bound to three Ti atoms. Ti13 · 6THF
˚
is a two-shell cluster with a Ti O distance of 2.02 A and
˚
a Ti Ti distance of 2.89 A. Fig. 1 shows the normalized Ti
K-edge XANES spectra of sodium alanate doped with Ti
clusters after ball milling (bm), at defined stages of the de-
hydrogenation (dd), hydrogenation (da) and after the second
absorption reaction (a2a). For comparison the spectra of the
Ti foil and the pure Ti13 · 6THF (Ti13) is added. From the
first derivatives of the spectra (not shown) it can be seen that
the first inflection point in spectra (Ti13), (bm) and (dd) is
at 4967 eV, whereas it is lowered to 4966.5 eV in the spectra
(da) and (a2a). This indicates that Ti in the clusters is trans-
formed from the coordinated (‘sub-oxide’) to the metallic
state during the hydrogenation reaction. First EXAFS analy-
sis confirms that in the case of the cluster the local structure
around Ti atoms is changing during the hydrogenation reac-
tion. Comparison of the Fourier Transform magnitude of the
EXAFS signal shows that the amplitude of the oxygen shell
is reduced during the hydrogenation and subsequent cycles.
This indicates that the Ti O bonds are stripped off from the
metal core during the first absorption reaction.
3. Results
3.1. State of the catalyst
The chemical state of Ti in NaAlH4 doped with 5 mol% of
either TiCl3 or Ti13 · 6THF was investigated using X-ray ab-
sorption spectroscopy and X-ray photoelectron spectroscopy.
3.1.1. TiCl3 doped samples
XANES [12] and XPS analysis [19] confirm earlier as-
sumptions of a reduction of the precursor TiCl3 occurring
during ball milling [4]. The titanium is partially reduced to
the metallic state after 2 min of ball milling and completely
reduced after 30 min of ball milling. Moreover, Ti remains
in a zero-valent state during desorption and absorption of
hydrogen. EXAFS analysis [12] suggests a local structure
around the Ti atoms consisting only of Ti neighbors after ball
milling and during subsequent release and uptake of hydro-
gen. Furthermore, the structure formed after ball milling is
in a distorted state but the local order evolves and tends to a
higher order from the ball-milled state to the absorbed one.
The coordination environment found from the refinement of
the EXAFS function χ(k) is close to that of metallic titanium.
3.2. Distribution of Ti in doped samples
It was suggested in earlier work that Ti would not only stay
at the surface, but might also migrate into the lattice, where it

734
M. Fichtner et al. / Journal of Alloys and Compounds 404–406 (2005) 732–737
could assist the phase transformation [13]. This presumption
was supported by theoretical work which showed that partial
replacement of Al by Ti may be beneficial from an energetic
point of view [14]. However, no experimental evidence has
been presented for the assumption yet.
tron density is not sufficient to describe properly some of the
Bragg reflections. In contrast to this, the refinement of the Na
occupation number revealed a full occupation of the site and
no change in the reflection description.
For these reasons, a Ti cation (up to 20 at.%) was randomly
substituted for Al to see whether the intensities description
improved. The total amount of Al + Ti was equal to the crys-
tallographic site occupation number. The two Rietveld cal-
culated patterns with and without Ti at the Al position inside
the lattice are compared in Fig. 3.
Even if the starting amount of Al and Ti corresponded to
a fully occupied site, the results obtained by the refinement
could indicate the presence of vacancies in the alanate struc-
ture. Both Ti and Al can be present but the global amount
is lower than the stoichiometric value. However, due to the
highcorrelationsbetweenthermalparametersandoccupation
numbers further investigations are in progress to determine
if Ti substitutes for Al and the exact amount of Ti, if present,
inside the lattice.
3.2.1. XRD analysis
Our approach was to check first whether indications of
a Ti/Al exchange could be found in X-ray diffraction data.
In order to investigate possible structural modifications in
doped sodium alanate, XRD patterns were taken from puri-
fied NaAlH4 which was ball-milled for 30 min either with
5 mol% of TiCl3 or with of 5 mol% Ti13 · 6THF.
NaAlH4 crystallizes in the tetragonal space group I41/a,
Z = 4, a = 0.5025 nm, c = 1.136 nm, Na in 4a (0, 1/4, 1/8),
Al in 4b (0, 1/4, 5/8), H in 16f (x, y, z). The H general po-
sition and the thermal parameters for Al, Na and H were
taken from neutron diffraction data [16]. Na3AlH6 crystal-
lizes in the monoclinic space group P21/n. Al is cubic, space
group Fm3m, a = 0.4049 nm. The XRD data were analyzed
by Rietveld refinement using the DBWS-9600 code written
by Sakthivel and Young [15].
3.2.1.2. TiCl3 doped samples. The XRD pattern was found
to be composed of the NaAlH4 phase, metallic Al, and some
traces of TiCl3. Na3AlH6 was not present. As for the sample
ball-milled with Ti clusters the NaAlH4 intensities were not
properly described using a pure sodium alanate phase. The
same refinement procedure used for the Ti cluster sample and
described in the preceding paragraph was applied here.
Also in this case, the results of the refinement showed that
the Al site could accommodate Ti atoms. Again, a high corre-
lation between thermal parameters and occupation numbers
was observed; the refinement of the thermal parameters with
the occupation numbers for the NaAlH4 phase without any
Ti inside being kept fixed did not result in any improvement
of the XRD data description. Only a low amount of vacancies
was found inside the lattice. This could probably be due to
the small quantity of Ti present in the lattice.
3.2.1.1. Ti cluster doped samples. The pattern contained
contributions coming from NaAlH4 phase, traces of the first
stage decomposition of the alanate Na3AlH6 and some metal-
lic aluminum. There were no indications of any preferred
orientation in the samples.
The results of the Rietveld refinement are reported in Fig.
2. The inset is the enlargement of the 2θ range from 17 to 22◦
showing the (0 1 1) reflection of the NaAlH4 phase.
It is clearly evident from both figures that some of the
alanate intensities are not properly described by the refine-
ment. The attempt of refining the thermal parameters did not
improve the intensities description. A refinement of the Al
occupation number (with the other occupation numbers and
all the thermal parameters kept fixed) showed that Al elec-
Fig. 3. Rietveld calculated pattern with Ti partly substituting Al and with-
out Ti in the lattice. The arrows indicate major deviations between the two
simulated patterns.
Fig. 2. Observed and calculated pattern of NaAlH4 doped with 2 mol% Ti on
Ti cluster basis. The arrows indicate major deviations between the observed
and calculated data.

M. Fichtner et al. / Journal of Alloys and Compounds 404–406 (2005) 732–737
735
3.2.2. Surface analysis
To investigate the spatial distribution of Ti in Ti doped
sodium alanate samples, surface analytical studies have been
performed. Secondary ion mass spectrometry (SIMS) was
used to determine chemical species at the outer surface of
doped alanate samples. Secondary neutral mass spectrometry
(SNMS) was applied for depth-profiling. SNMS is a surface
analysis method which detects secondary neutrals emitted
from a sputtered surface and allows for the semi-quantitative
determination of atom concentrations also of non-conducting
powder materials with nanometer-resolution [17]. The yield
of sputtered secondary ions strongly depend on the ionization
probability at the actual surface. Therefore, SIMS cannot be
used to quantify species with a varying surface composition.
However, it gives information on the elemental composition,
and, as sputtered cluster ions carry information about neigh-
boring atoms on a samples surface, the method can be used
for chemical speciation of complex salts [18].
Fig. 4. SIMS spectrum of the outer surface of sodium alanate doped with
4 mol% TiCl3 and ball-milled for 300 min.
of Na (m/z = 23), Al (m/z = 27), Cl (m/z = 350), Ti
(m/z = 48), and In (m/z = 115) were recorded as a
function of time. Fig. 5 shows the Ti intensities for the three
differently milled samples as well as for a sample prepared
by mixing ball-milled alanate with 4 mol% TiCl3 in an agate
mortar. The data exhibits a development of the Ti intensity
with depth, depending on the milling time. The sample
which was ball-milled for 3 min showed a maximum of the
Ti concentration in the outer region. The data suggests that a
Ti rich layer forms at the outer surface of the alanate particles
during short-term ball-milling. From the depth propagation
values mentioned above, the thickness of this layer can be
estimated to be some 30–40 nm. The data also show that the
layer vanishes and Ti is dispersed in the material when the
sample is ball-milled for 30 min or more.
3.2.2.1. SIMS measurements. SIMS spectra of positive sec-
ondary ions were taken from a fresh sample which was pre-
pared by ball milling of NaAlH4 with 4 mol% TiCl3 for
300 min. Fig. 4 shows the mass spectrum with the most in-
tense signals of Na+ (m/z = 23) and Al+ (m/z = 27) com-
ing from the alanate. K+ (m/z = 39, 41) is a typical trace
found in sodium containing salts, In+ (m/z = 115, 113) was
emitted from the embedding In foil. Surprisingly, the Ti sig-
nal (Ti+ (m/z = 46, 47, 48) is comparably low at the surface
of the composite particles. The spectrum rather shows intense
cluster signals coming from Na+2 (m/z = 46, normally ob-
served with high Na+ intensities) and Na2O+ (m/z = 62).
The NaAl+ (m/z = 50), NaAlO+ (m/z = 66) and Na2AlO+
(m/z = 89) signals suggest Na and Al in close vicinity, which
is probably due to an alanate species. The oxygen in the clus-
ter ions indicates a partly oxidized alanate species at the sur-
face, due to a strong interaction of ball-milled alanate even
with traces of oxygen. Moreover, the strong signal of the
Na2Cl+ cluster ion is a direct proof of the precursor TiCl3
having reacted with sodium alanate to form NaCl.
Obviously, the Ti concentration is reduced at the outer
surface of samples that have been ball-milled for 30 and
300 min, which is accordance with the SIMS data presented
3.2.2.2. SNMS measurements. It was not possible to deter-
mine the depth propagation and resolution of alanate samples
as standards are lacking. However, a rough estimation is pos-
sible by comparison with data obtained with a layer pigment
(Iriodin 9444, MERCK). The pigment particles consist of a
mica core coated with a 170 nm TiO2 layer followed by a
35 nm Cr2O3 layer.
It took 1000–1100 s to sputter through the Cr2O3
layer which corresponds to a depth propagation of 0.03–
0.035 nm/s. In the second layer, the depth propagation was
0.04–0.045 nm/s. The time difference between the 33% and
the 67% intensity value of the increasing Ti signal after sput-
teringthefirstlayerwastakentocalculatethedepthresolution
which was 8–9 nm in this particular sample.
Fig. 5. SNMS time profile of sodium alanate doped with 4 mol% TiCl3 and
ball-milled for 3, 30, and 300 min or loosely mixed. The shown Ti intensities
were normalized for comparison.
Doped alanate samples were prepared by ball-milling
NaAlH4 with 4 mol% TiCl3 for 3, 30, and 300 min. Intensities

736
M. Fichtner et al. / Journal of Alloys and Compounds 404–406 (2005) 732–737
rehydrogenation was found to be limited by diffusion of two
solid phases, which have to react with each other under the
presence of hydrogen.
Hence, we assume that Ti has an influence also on the rate
of these processes. The results presented in Section 3.2 in-
dicate that Ti atoms may be driven into the material by ball
milling for 30 min or longer. Moreover, the fact that Ti is
dispersed in the material correlates with the rate of hydrogen
exchange, which corresponds to the rate of phase transforma-
tion. If chemisorption at a Ti containing phase in the surface
controlled the rate of transformation, the reduced Ti concen-
tration at the outer surface in case of longer milling times
would favor a contradictory trend of the data. It is therefore
unlikely that the Ti concentration at the outer surface has a
dominating influence on the catalytic effect.
The XRD study in Section 3.2 suggests that a part of the Ti
can also substitute Al in the alanate lattice. There, it can cre-
ate non-thermal defects due to its different size, depending on
the valence state, and coordination behavior when compared
to Al. Defects are known to enhance diffusion extraordinar-
ily and hence play a determining role for materials transport
in many solids. From the findings mentioned above, we as-
sume that Ti substituting Al in the alanate lattice is at least
partly responsible for the enhanced transformation rates. The
presumption is supported by the observation that Ti clusters
create more vacancies in the alanate than TiCl3. This could
explain the higher transformation rates of Ti cluster-doped
samples.
Fig. 6. Isothermal decomposition of NaAlH4 + 2 mol% TiCl3 at 150 ◦C, in
the third desorption–absorption cycle (experimental details described in [7]).
above. This is also confirmed by XPS measurements of doped
sodium alanate, where a reduced Ti concentration was found
at the surface of samples that had been subjected to longer
ball milling [19]. The findings support the assumption that Ti
is driven into the material by high energy ball milling.
3.3. Phase transformations
Phase separation is observed when the alanate is thermally
decomposed into Na3AlH6 and Al or NaH and Al, depending
on the stage of decomposition [1,13]. Taking low concentra-
tions of the Ti cluster as dopant, kinetic studies showed that
nucleation and growth of the forming phases are rate limit-
ing for the transformations during hydrogen exchange for the
observed temperature range of 100– 150 ◦C [20]. The study
indicates that diffusion of the involved species (NaH + Al and
Na3AlH6 + Al) limits the alanate formation rate during the
rehydrogenation of the decomposed material [20].
Fig. 6 shows the kinetic data for the decomposition of
doped NaAlH4 which was ball-milled for various times. It
is evident that samples ball-milled for 3 min exhibit the low-
est transformation rates, whereas all other samples transform
considerably faster with about the same rates. The corre-
sponding absorption kinetics are not shown here, but exhibit
the same behavior.
Moreover, if the whole transformation process is consid-
ered a consecutive reaction of a fast chemisorption step and
a rate-determining, defect-enhanced transformation in the
solid, it becomes clear why other dopants with much bet-
ter chemisorption properties hardly have any catalytic effect.
Although they may increase the rate of the chemisorption
step, the rate of the second, rate-determining step, is low be-
cause of the lack of defects created by the substitution of Al
in the lattice.
5. Summary
The chemical state and spatial distribution of Ti in doped
sodium alanate samples was investigated by a combination
of X-ray absorption (XANES, EXAFS), X-ray diffraction
(XRPD) and surface analytical (SIMS, SNMS, XPS [19])
methods.
4. Discussion
XANESandEXAFSdataindicatesthatTiCl3 isreducedto
Ti0 during the ball-milling process and stays in this state dur-
ing desorption and absorption of hydrogen. The next neigh-
bors of Ti in freshly ball-milled, in partially dehydrogenated
and in partially rehydrogenated samples were determined to
be at typical Ti Ti distances and no hint for the formation
of an alloy with Al could be found at this early stage of the
process. X-ray diffraction data and Rietveld refinement on
NaAlH4 samples doped with Ti clusters and TiCl3 suggest
the incorporation of Ti into the alanate lattice where it can
The Ti catalyst has a large influence on the rate of H re-
lease and uptake and the role played by Ti in the materials
transformations is of great importance. It is expected that
a Ti containing phase may act as chemisorption site for the
molecular hydrogen, which was also demonstrated in isotope
scrambling experiments [21].
However, the rate limiting step observed is governed
by nucleation and growth as was shown in kinetic stud-
ies. The formation rate of the alanate phase during the

M. Fichtner et al. / Journal of Alloys and Compounds 404–406 (2005) 732–737
737
stay at Al sites. The data also indicate the formation of va-
cancies in the material.
[6] J. Chen, N. Kuriyama, Q. Xu, H.T. Takeshita, T. Sakai, J. Phys. Chem.
B 105 (2001) 11214.
[7] M. Fichtner, O. Fuhr, O. Kircher, J. Rothe, Nanotechnology 14 (2003)
778.
[8] B. Bogdanovic, M. Felderhoff, S. Kaskel, A. Pommerin, K. Schlichte,
F. Schu¨th, Adv. Mater. 15 (2003) 1012.
[9] M. Fichtner, O. Fuhr, O. Huebner, O. Kircher, W. Klopper, A. Le´on,
Proceedings of the Seventh International Conference on Nanostruc-
tured Materials, NANO 2004, Wiesbaden, Germany, June 20–24,
2002.
[10] K. Hosokawa, Thesis, Iowa State University, Ames, IA, 2002.
[11] M. Fichtner, O. Fuhr, J. Alloys Compd. 345 (2002) 286.
[12] A. Le´on, O. Kircher, J. Rothe, M. Fichtner, J. Phys. Chem. B 108 (2004)
16372.
[13] K.J. Gross, S. Guthrie, S. Takara, G. Thomas, J. Alloys Compd. 297
(2000) 270.
XPS and SNMS data of TiCl3 doped samples exhibit an
initiallyformedTirichlayeratthesurfacewhichisdistributed
in the material when the material is ball-milled for 30 min or
longer. The remaining surface region appears to be depleted
of Ti, which is distributed in the remaining material, accord-
ing to SNMS and XPS data.
The experimental observations and correlations support a
mechanism where the number of defects created by a partial
substitution of Al by Ti determines the rates of transformation
of the alanate material.
[14] O.M. Løvvik, S.M. Opalka, Phys. Rev. B 71 (2005) 54103.
[15] A. Sakhtivel, R.A. Young, User’s Guide to Programs DBWS-9OO6PC,
Georgia Institute of Technology, Atlanta, GA, USA, 1990.
[16] B.C. Hauback, H.W. Brinks, C.M. Jensen, K. Murphy, A.J. Maeland,
J. Alloys Compd. 358 (2003) 142.
[17] J. Goschnick, C. Natzeck, M. Sommer, Appl. Surf. Sci. 144–145 (1999)
31.
[18] M. Fichtner, J. Goschnick, H.J. Ache, Fresenius J. Anal. Chem. 348
(1994) 201.
[19] A. Le´on, D. Schild, M. Fichtner, J. Alloys Compd., this issue.
[20] O. Kircher, M. Fichtner, J. Appl. Phys. 95 (13) (2004)
7748.
[21] F. Schu¨th, B. Bogdanovic, M. Felderhoff, Chem. Commun. 20 (2004)
2249.
References
[1] (a) B. Bogdanovic, M. Schwickardi, J. Alloys Compd. 253–254 (1997)
1.
(b) B. Bogdanovic, R.A. Brand, A. Marjanovic, M. Schwickardi, J.
To¨lle, J. Alloys Compd. 302 (2000) 36.
[2] A. Zaluska, L. Zaluski, J.O. Stro¨m-Olsen, J. Alloys Compd. 298 (2000)
125.
[3] C.M. Jensen, K.J. Gross, Appl. Phys. A 72 (2) (2001) 213.
[4] K.J. Gross, G. Sandrock, G.J. Thomas, J. Alloys Compd. 330–332
(2002) 691.
[5] R.A. Zidan, S. Takara, A.G. Hee, C.M. Jensen, J. Alloys Compd. 285
(1999) 119.