Investigation of hydrogen discharging and recharging processes of Ti-doped NaAlH4 by X-ray diffraction analysis (XRD) and solid-state NMR spectroscopy

Article abstract of DOI:10.1016/S0925-8388(02)00953-2

The processes occurring in the course of two sequential hydrogen discharging and recharging cycles of Ti-doped sodium alanate were investigated in parallel using XRD analysis and solid-state NMR spectroscopy. Both methods demonstrate that in hydrogen storage cycles (Eq. (1)) the majority phases involved are NaAlH₄, Na₃AlH₆, Al and NaH. Only traces of other, as yet unidentified phases are observed, one of which has been tentatively assigned to an Al-Ti alloy on the basis of XRD analysis. The unsatisfactory hydrogen storage capacities heretofore observed in cycle tests are shown to be due entirely to the reaction of Na₃AlH₆ with Al and hydrogen to NaAlH₄ (Eq. (1), 2nd hydrogenation step) being incomplete. Using XRD and NMR methods it has been shown that a higher level of rehydrogenation can be achieved by adding an excess of Al powder.

Full text of DOI:10.1016/S0925-8388(02)00953-2

Journal of Alloys and Compounds 350 (2003) 246–255
L
www.elsevier.com/locate/jallcom
Investigation of hydrogen discharging and recharging processes of
Ti-doped NaAlH4 by X-ray diffraction analysis (XRD) and
solid-state NMR spectroscopy
*
¨
´
¨
B. Bogdanovic, M. Felderhoff, M. Germann, M. Hartel, A. Pommerin, F. Schuth , C. Weidenthaler,
B. Zibrowius
¨
¨
Max-Planck-Institut fur Kohlenforschung, Kaiser-Wilhelm-Platz 1, D-45470 Mulheim an der Ruhr, Germany
Received 17 April 2002; received in revised form 5 August 2002; accepted 5 August 2002
Abstract
The processes occurring in the course of two sequential hydrogen discharging and recharging cycles of Ti-doped sodium alanate were
investigated in parallel using XRD analysis and solid-state NMR spectroscopy. Both methods demonstrate that in hydrogen storage cycles
(Eq. (1)) the majority phases involved are NaAlH4, Na3AlH6, Al and NaH. Only traces of other, as yet unidentified phases are observed,
one of which has been tentatively assigned to an Al–Ti alloy on the basis of XRD analysis. The unsatisfactory hydrogen storage
capacities heretofore observed in cycle tests are shown to be due entirely to the reaction of Na3AlH6 with Al and hydrogen to NaAlH4
(Eq. (1), 2nd hydrogenation step) being incomplete. Using XRD and NMR methods it has been shown that a higher level of
rehydrogenation can be achieved by adding an excess of Al powder.
2002 Elsevier Science B.V. All rights reserved.
Keywords: Hydrogen absorbing materials; Metal hydrides; Gas–solid reaction; Nuclear resonances
1. Introduction
therefore to investigate the causes for this detrimental
effect by means of XRD analysis and solid state NMR
spectroscopy. In the course of these investigations, both
methods proved to be excellent tools for providing in-
formation about the processes occurring in the material on
a microscopic level during hydrogen discharging and
recharging. While an in situ X-ray diffraction study of the
decomposition of NaAlH4 has already been published by
Gross et al. [7], solid-state NMR spectroscopy has very
recently been used by Balema et al. [8] to study the effect
of ball-milling on LiAlH4.
One of the main goals associated with the development
of Ti (or other metal)-doped NaAlH4 as hydrogen storage
materials [1–6] is to achieve the highest possible hydrogen
storage capacity in a large number of cycles. The calcu-
lated hydrogen storage capacity of neat NaAlH4 (Eq. (1))
is 5.5 wt.% H2. However, in practice the capacity is
always lower, due to the added weight of the dopant and,
in some cases, also due to the consumption of a fraction of
NaAlH4 in the doping reaction. In general, the lower the
dopant concentration the higher is the maximum storage
capacity. If the stoichiometry of the doping reaction is
known, the theoretically achievable storage capacity can be
calculated.
2. Experimental
A major disadvantage of the NaAlH4 storage materials
investigated previously is that storage capacities in cycle
tests were always lower than the theoretically expected
capacity. One of the primary goals of the present work was
The X-ray powder patterns for qualitative and quantita-
tive phase analysis were collected on a Stoe STADI P
transmission diffractometer with a primary monochromator
and a linear position sensitive detector. The data were
collected in the range between 158 and 1208 2u with a step
width of 0.018 2u. The samples were filled in a glove box
into glass capillaries ([ 0.5 mm) which were sealed to
prevent contact with air during the measurement. The
*
Corresponding author. Tel.: 149-208-306-2373; fax: 149-208-306-
2995.
E-mail address: schueth@mpi-muelheim.mpg.de (F. Schuth).
¨
0925-8388/02/$ – see front matter
PII: S0925-8388(02)00953-2
2002 Elsevier Science B.V. All rights reserved.

´
B. Bogdanovic et al. / Journal of Alloys and Compounds 350 (2003) 246–255
247
recorded patterns were evaluated qualitatively by com-
parison with entries from the PDF-2 powder pattern
database and theoretical patterns calculated from literature
data.
The quantitative phase analysis was performed by
Rietveld refinements using the program FullProf2000 by
ture rise resulting in a sudden release of hydrogen.) All the
dehydrogenated samples caused a strong spinning rate-
dependent de-tuning of the probe.
The spectra were recorded using single pulse excitation
employing short pulses and small flip angles (23Na: 0.9 ms,
p/8; 27Al: 0.6 ms, p/12). The 23Na and 27Al NMR spectra
were referenced with respect to aqueous solutions of
sodium chloride and aluminum nitrate, respectively.
´
Rodrıguez-Carvajal [9]. Phases which could not be iden-
tified (their concentration is estimated to be well below 5
wt.%) were neglected during the refinement and therefore
only the ratios of the analyzed crystalline phases are
correctly determined, but not their absolute values.
Another source of error is the amount of amorphous phase.
The quantity of amorphous phase has to be determined by
the addition of an exactly defined amount of internal
standard. Due to the sample handling under protective gas,
the admixture and homogenization of an internal standard
to a defined amount of sample is not trivial. If the
preparation of sample and standard is not exact, the
analysis of the amorphous phase will have a large error.
Therefore, no quantification of the amorphous fraction has
been attempted. The content of phases at different stages of
a cycle test as determined by Rietveld analysis is given in
Table 1.
NMR measurements were performed on a Bruker
Avance 500WB spectrometer operating at a magnetic field
of 11.74 T. The corresponding resonance frequencies of
23Na and 27Al are 132.26 and 130.29 MHz, respectively. A
4-mm MAS probehead with dry nitrogen as driving and
bearing gas was used. The rotors were filled in a glove box
and stored under argon prior to the measurements. Spin-
ning rates of up to 15 kHz were used. In the case of some
of the dehydrogenated samples, eddy currents generated in
metallic particles in the sample prevented us from using
high spinning rates. (Warning: an attempt to spin such
samples at higher rates can lead to a substantial tempera-
2.1. Preparation of samples of Ti-doped NaAlH4
All experiments were carried out under an argon atmos-
phere, using a glove box or the Schlenk technique.
Experiment A: 4.06 g (75 mmol) of recrystallized
NaAlH4 [2] were suspended in 50 ml of toluene and 0.51 g
(1.5 mmol) of Ti(OC4H9n)4 were added to the stirred
suspension using a syringe. The hydrogen evolution taking
place during stirring of the suspension at room temperature
was recorded by means of a thermovolumetric apparatus
[1]. The hydrogen evolution took place with constant rate
and stopped abruptly after 70 h giving 78 ml of H2 (1.9
mol H2 /mol Ti). The black suspension was filtered, the
solid washed with toluene and pentane and dried in
vacuum until the constant weight (4.40 g) was achieved.
A sample of Ti-doped NaAlH4 (1.10 g) was heated in a
thermovolumetric apparatus at first to 120 8C and then to
180 8C (Fig. 1) and the hydrogen evolution taking place at
the same time (—) was recorded. The heating was stopped
for a short time and the glass vessel taken out of the oven
to remove small samples for XRD and NMR measure-
ments. These events are marked by circles in Fig. 1.
Hydrogenations were carried out at 100 8C/100 bar for 12
h in a 35-ml autoclave. Hydrogen capacities (Fig. 1): (1)
cycle, 4.5 wt.%; (2) cycle 4.0 wt.% H2.
Experiment B: In a parallel experiment, a sample of
Table 1
Quantitative Rietveld analyses of the phase contents in the course of dehydrogenation and rehydrogenation of Ti-doped NaAlH4
Sample
code
Cycle
number
Stage of the cycle
testa
Phase content (mol%)
Al/Na3AlH6 ratio
NaAlH4
Na3AlH6
Al
NaH
By-
product
From
XRD
From 27Al
MASNMRc
A0
A1
1
1
Starting material
100
49
2
2
2
2
2
1
2
2
Dehydrogenated, about
50% of the 1st step
Dehydrogenated, 1st step
Dehydrogenated, 2nd step
Rehydrogenatedb
15
36
2.4
1.0
A2
A3
A4
A5
A6
B4
B5
1
1
1
2
2
1
1
7
2
30
1
3
2
15
16
9
63
46
51
44
24
37
21
2
53
2
56
2
4
1
1
1
1
1
1
1
2.1
46
17
1.3
2
46
2
0.9
2
Dehydrogenated, 2nd step
2
Rehydrogenatedb
61
43
70
1.6
2.3
2.3
2
Rehydrogenatedb
1.4
0.2
Milled and
2
rehydrogenatedb
B6
1
Al added to B4, milled
80
2
20
2
2
2
2
and rehydrogenatedb
a The cycle test is represented in Fig. 1.
b Hydrogenation conditions: 100 8C/100 bar/12 h.
c The effects of final bandwidths of excitation and probe are neglected (see text).

´
B. Bogdanovic et al. / Journal of Alloys and Compounds 350 (2003) 246–255
248
Fig. 1. De- and rehydrogenation cycles carried out with a sample of Ti-doped NaAlH4 (2 mol%). The circles with the corresponding numbers indicate the
stages at which samples were taken for XRD and NMR analyses. Solid line, hydrogen evolution; dotted line, sample temperature; dashed line,
rehydrogenation step.
Ti-doped NaAlH4 was subjected to one dehydrogenation/
rehydrogenation cycle in the same way as in experiment A.
To 0.44 g of the rehydrogenated sample (B4) 65 mg (2.4
mmol) of Al grindings (Al bronze, ‘Lunasol’, Schop-
flocher, Frankfurt/M.; 12 m2 /g) were added in a glove
box. The mixture was ball-milled for 30 min/30 Hz in a
10-ml steel vessel together with two high-grade steel balls
(Ø512 mm) using a Retsch high energy ball mill. There-
after the sample was hydrogenated at 113 8C/135 bar for
20 h. Blank test (B5): Repeated ball-milling and hydro-
genation of sample B4 without addition of Al powder at
the same conditions as for B6.
rehydrogenation processes of Ti-doped NaAlH4 according
to Eq. (1).
NaAlH4 1. dehydroágenation step 1/3Na3AlH6 1 2/3 Al 1
2. hydrogenation step
H2á2. dehydrogenation step
á
NaH 1 Al 1 3/2H2
(1)
1. hydrogenation step
Fig. 2 shows a part of the powder diagram of the starting
material (A0) after the final Rietveld run. For the refine-
ment of the structure of NaAlH4, the crystal structure data
of Lauher et al. [10] were used. All the observed reflec-
tions belong to NaAlH4. The reported [11] changes of the
crystal lattice parameters of Ti-doped NaAlH4 are not
observed in our study. Two additional series of samples
with different concentrations of the Ti catalyst (1 and 3
mol.%) were prepared and investigated by X-ray analysis.
An increase or decrease in the lattice constants for the
NaAlH4 phase cannot be confirmed in any of the samples.
The 23Na and 27Al NMR spectra of Ti-doped sodium
alanate at different stages of dehydrogenation (cf. Table 1)
are given in Figs. 3 and 4, respectively. The high quality of
the sodium alanate used as starting material in combination
with the comparably high magnetic field and the spinning
speeds applied allows the spectral parameters (chemical
3. Results and discussion
The samples of Ti-doped NaAlH4 used for XRD and
NMR investigations were prepared by doping of NaAlH4
with Ti(OC4Hn9)4 (2 mol%) in toluene [2]. One of the
doped samples was subjected to two complete dehydroge-
nation/rehydrogenation cycles (Fig. 1). The taking of
samples at different stages of the cycle test is marked by
circles and the samples labeled as A0–A6. In Table 1 the
samples A0–A6 are assigned to dehydrogenation and

´
B. Bogdanovic et al. / Journal of Alloys and Compounds 350 (2003) 246–255
249
Fig. 2. Powder pattern of Ti-doped NaAlH4 after the final Rietveld run. The observed data points (cross), the calculated points (solid line), and the
difference curve are given. The peak positions are marked by lines in the lower part of the pattern.
shift, quadrupole coupling) for NaAlH4, Na3AlH6 and
NaH to be determined with an accuracy and reliability
superior to all previously published data [12,13]. A more
detailed presentation of the NMR results will be published
in a forthcoming paper (B. Zibrowius, unpublished). For
identifying and monitoring the different compounds in the
course of dehydrogenation/rehydrogenation cycles the
positions of the resonance lines are sufficient. These line
positions are summarized in Table 2.
The dehydrogenation to approximately half of the first
dissociation step (A1) leads to the formation of Na3AlH6
beside aluminum. The main crystalline phase detected by
XRD at this stage is NaAlH4, followed by Na3AlH6, and
Al, respectively (Fig. 5). The crystal structure data re-
the XRD data: present at this stage are NaAlH4, Na3AlH6
and Al.
The XRD (Fig. 5, A2) indicates that after dehydrogena-
tion at 120 8C for 4 h the sample still contains some small
amount of NaAlH4, while aluminum and Na3AlH6 are the
main crystalline phases (Table 1). The impurity phase
observed for the sample A1 is still present. The 23Na NMR
spectrum of this sample (Fig. 3, A2) shows no NaH at all,
but only Na3AlH6. Because of the small difference in the
line positions observed for NaAlH4 and the Na(2) position
of Na3AlH6 (Table 2), it would be difficult to identify the
small amount of NaAlH4 detected by XRD in the 23Na
MAS NMR spectrum. However, the 27Al MAS NMR
spectrum (A2, Fig. 4) shows a line with low intensity
corresponding to NaAlH4, confirming the XRD data. The
first stage of the dehydrogenation process (Eq. (1)) is thus
almost complete at this stage. Complete dehydrogenation
could certainly be achieved by prolonging the heating
under these conditions.
¨
ported by Ronnebro et al. [14] were used for the simulation
of Na3AlH6. At 12.98 and 20.68 2u (Fig. 5) two reflections
are observed which cannot be assigned to any of the
identified phases. These reflections are considered to
belong to impurities with no relevance to the overall
process because they were detected only in some of the
experiments which were otherwise identical to the others
with respect to the kinetics and storage capacities. NMR
data (traces A1 in Figs. 3 and 4) are in full agreement with
In the completely dehydrogenated sample, NaH could be
detected as a new phase both by XRD (A3, Fig. 5) and
NMR spectroscopy (A3, Fig. 3). For Rietveld analysis, the
structure data of NaH given by Shull et al. [15] were used.

´
B. Bogdanovic et al. / Journal of Alloys and Compounds 350 (2003) 246–255
250
Fig. 4. 27Al MAS NMR spectra (9600 scans, 5 s recycle delay) of the
samples A0–A4 (Table 1). The spinning rates applied are 15 kHz (A0),
13.5 kHz (A1 and A4), 12 kHz (A2), and 11 kHz (A3).
Fig. 3. 23Na MAS NMR spectra (240 scans, 15 s recycle delay) of the
samples A0–A4 (Table 1). The spinning rates applied are 15 kHz (A0,
A1, and A4), 12 kHz (A2), and 10 kHz (A3).
sodium, below the detection limit of the X-ray diffraction).
Careful integration of the intensities of the central transi-
tion (excluding contributions from overlap with satellite
transition spinning side bands but including side-band
intensities from the central transition) of the 23Na and 27Al
MAS NMR spectra allowed the molar ratio of Na3AlH6 to
NaAlH4 to be estimated as 0.2760.03. Since there is
almost no NaH left in the sample, according to Eq. (1) the
amount of metallic aluminum should be twice the amount
of Na3AlH6. This intensity ratio was not observed for the
resonance lines in the 27Al MAS NMR spectrum of A4
The content of aluminum at this stage (Table 1, A3) is
almost equal to that of NaH, as required by the stoi-
chiometry, while the content of Na3AlH6 is almost negli-
gible. The unidentified phase observed for the samples A1
and A2 is still present. The existence of a by-product can
also be inferred from 27Al MAS NMR spectroscopy. The
additional line at about 270 ppm (marked with asterisks in
the right column of Fig. 4) is probably caused by some
aluminum oxide introduced into the parent material A0 by
the aluminum source. Since the aluminum used in the
synthesis of the materials is a very fine particle grade, the
surface aluminum oxide present can be detected in the
NMR experiments. The intensity of the resonance line of
this by-product varied with the batch of sodium alanate,
corresponding to different amounts of surface oxide pres-
ent in the starting material.
Table 2
Positions of the NMR lines (in ppm) of the compounds detected in
dehydrogenated sodium alanate
Compound
23Na
27Al
The rehydrogenated sample (Table 1, A4) contains
NaAlH4 as the main component (based on weight) besides
aluminum and Na3AlH6. NMR spectroscopy also detected
primarily Na3AlH6, NaAlH4 and metallic aluminum, to-
gether with a small amount of NaH (less than 3% of the
NaAlH4
Na3AlH6
29.460.2
Na(1): 22.860.2
Na(2): 210.260.2
18.360.2
95.260.5
242.760.3
NaH
Almet
–
–
1639.560.5

´
B. Bogdanovic et al. / Journal of Alloys and Compounds 350 (2003) 246–255
251
Fig. 5. Comparison of the powder patterns of the analysed samples A0–A6, the reflections of the crystalline phases are marked by symbols. The reflections
which could not be assigned to any known phase are marked by ‘?’.
(Fig. 4), but for quantitative analysis one has to take into
account that the cubic structure of metallic aluminum leads
to vanishing quadrupolar coupling. Therefore, the reso-
nance line of metallic aluminum contains the contributions
of all the five possible transitions [16]. To compare its
intensity with that of the central transition observed for
Na3AlH6, it has to be multiplied by a factor of 9/35. The
thus derived Almet /Na3AlH6 ratio for sample A4 is 0.86.
Even smaller values have been found by NMR spec-
troscopy for other rehydrogenated samples (cf. Table 1).
By shifting the frequency offset and tuning to the reso-
nance line of the metallic aluminum, it was found that only
part of the missing signal intensity for metallic aluminum
in the spectrum of sample A4 can be explained by unequal
rf excitation and optimization of the probe tuning for the
central transition of NaAlH4.
The skin effect occurring in larger metallic particles
present in the sample and an inhomogeneous distribution
of these metallic particles in the batch are very probably
the reason for the discrepancy between the expected and
observed Almet /Na3AlH6. The fast sample spinning itself
might have led to a further aggregation of the rather large
metallic aluminum particles (several 100 nm in size)
detected by electron microscopy. At a spinning rate of 10
kHz the acceleration exerted at the inner wall of a 4-mm
MAS rotor is about 600 000 times the standard accelera-
tion of gravity g. Spinning the sample at several kHz was
always found to increase irreversibly the de-tuning effect
observed for the dehydrogenated samples. As noted in the
experimental part, the generation of eddy currents was so
pronounced for some samples that spinning at high rates
was impossible and appreciable heating of the sample
occurred. Hence, the application of the calibration pro-
cedure used by Balema et al. [8] to quantify the amount of
aluminum in ball-milled LiAlH4 would not solve the
problem.
The ratio of Na3AlH6 to aluminum as determined from
the Rietveld analysis also does not correspond to the ratio
of 1:2 expected from the stoichiometry, and the NaAlH4 /
Na3AlH6 ratio determined from the XRD does not agree
with the one calculated from the NMR results. This is
attributed to the formation of amorphous or very small
domain size Na3AlH6 which cannot reliably be refined
using the procedure applied here (see Section 2). In other
samples or subsequent cycles (A6, for instance) such
discrepancies did not occur. The ratio determined by
Rietveld analysis of 15/24 for sample A6 corresponds to
1:2 within the error of the determination. These problems
only occurred with some of the samples, but they are
discussed in some detail here to highlight the problems in
analyzing the composition of such complex and highly
inhomogeneous samples. Thus, two independent methods
such as XRD and NMR spectroscopy should be used to
fully analyze the processes during cycling such hydrides.
If the cycle is then repeated, the same sequence of phase
transformations is essentially observed again, demonstra-
ting the complete reversibility of the process. The only
exception is the ratio of the different phases in the
rehydrogenated sample, as already discussed above. In
addition, kinetics and thus extent of rehydrogenation are
improved, as is almost invariably observed for titanium-
doped NaAlH4 in the second and subsequent dehydrogena-
tion/rehydrogenation cycles. This is tentatively interpreted
as being due to improved dispersion of the titanium on the

´
B. Bogdanovic et al. / Journal of Alloys and Compounds 350 (2003) 246–255
252
occurrence of eddy currents in the NMR experiment and
SEM investigations (not shown). During rehydrogenation
metallic aluminum particles would first react with NaH to
Na3AlH6, and in parallel Na3AlH6 and the residual
aluminum would react to NaAlH4. The metallic aluminum
particles would thus be coated by a layer of NaAlH4.
Those aluminum particles, which are rather big after
dehydrogenation, would after some period of rehydrogena-
tion consist of a center made of aluminum-coated with
NaAlH4. The aluminum in the particles core might then
not be reached by the reactant Na3AlH6 formed in other
parts of the sample, terminating the rehydrogenation
reaction.
In order to test this hypothesis additional experiments
were carried out. A regular sample was subjected to one
dehydrogenation/rehydrogenation cycle. After this cycle,
the powder pattern of the sample (Fig. 7, B4) again shows
the presence of NaAlH4, Na3AlH6, Al and NaH. After
addition of a fine Al powder to the sample B4, the mixed
sample was ball-milled for 30 min and again hydrogenated
(B6). The XRD pattern of the sample thereafter (Fig. 7,
B6) shows the complete absence of Na3AlH6 reflections.
A reference experiment, where the sample B4 was just
ball-milled and then exposed to a hydrogenation step
without additional aluminum, shows only a small reduction
of the amount of Na3AlH6 (B5). Hence, an excess of
aluminum in the system results in complete rehydrogena-
tion in the 2nd step, so that the full storage capacity can be
exploited. Normalized to weight, the additional aluminum
results in a lower overall storage capacity. Nevertheless,
the results indicate that improved mass transfer in the
system will result in a higher reversible hydrogen storage
capacity. The pattern of the refinement after the final
Rietveld run is given in Fig. 8.
The beneficial effect of excess aluminum on the rehy-
drogenation is also corroborated by the 23Na MAS NMR
spectra in Fig. 9. The spectrum of the dehydrogenated and
subsequently rehydrogenated sample B4 shows a signifi-
cant amount of Na3AlH6. While ball-milling and a further
rehydrogenation (B5) diminished the Na3AlH6 line already
significantly, the addition of excess aluminum prior to the
ball-milling and rehydrogenation under the same con-
ditions (B6) resulted in an almost complete conversion of
the intermediate Na3AlH6 into NaAlH4. The same conclu-
sion can also be drawn from the 27Al MAS NMR spectra
in Fig. 10.
Fig. 6. Powder pattern of sample A4. Aluminum reflection marked by the
bar, exhibiting a clear asymmetry at the right side of the reflection.
atomic scale, since the dehydrogenation/rehydrogenation
cycle is accompanied by a massive transfer of matter.
A possibly important effect is observed if one analyzes
the XRD patterns carefully. The aluminum reflections of
the rehydrogenated sample (A4) show a clear asymmetry
on the side of higher u values as an enlarged part of the
powder pattern in Fig. 6 shows. The powder patterns of the
samples with low aluminum content show that the
asymmetry is actually a shoulder which might be assigned
to the presence of another crystalline phase. It has been
reported for LiAlH4 that a titanium-catalyzed transforma-
tion to Al3Ti alloy occurs upon ball-milling with TiCl4 [8].
The shoulders on the right side of the aluminum reflections
observed in this work does not match the reflections of the
Al3Ti phase or any other known Al–Ti alloy. However,
since the intensity of the reflections increases with the
amount of the applied catalyst, and since alloys like Al2Ti
[17] have their main reflections at 2u values where the
broad shoulders appear, we suggest that a crystalline Al–
Ti alloy has formed during the phase transformation
processes. The nature of this phase is at present unclear,
but because of its possible relevance to the catalytic effect
of the titanium, the formation of an Al–Ti phase during
cycling of Ti-doped NaAlH4 will be further investigated.
One of the main results of the present XRD and NMR
investigations is that in the rehydrogenated samples A4
and A6 (Table 1 and Figs. 3–5) NaAlH4, Na3AlH6 and Al,
but only negligible amounts of NaH, could be observed.
This means that the rehydrogenation of NaH with Al to
Na3AlH6 (Eq. (1), 1st step) runs to completion, while the
reaction of Na3AlH6, Al and hydrogen to NaAlH4 (2nd
step) stops before it is completed, at least on the time scale
of the experiments.
4. Conclusion
The incomplete rehydrogenation of the Na3AlH6 is thus
the reason for low storage capacities in cycle tests (cf.
Section 1). A possible scenario could be a mass transfer
problem: during dehydrogenation rather large metallic
aluminum particles are formed, as evidenced by the
On the basis of the results presented above, the follow-
ing statements can be made with respect to hydrogen
discharging and recharging processes of Ti-doped NaAlH4:
1. XRD analysis and solid-state NMR spectroscopy are

´
B. Bogdanovic et al. / Journal of Alloys and Compounds 350 (2003) 246–255
253
Fig. 7. Powder patterns of the rehydrogenated sample B4, of the milled and rehydrogenated sample B5 and of the sample with the admixed aluminum B6.
The different phases observed in the two samples are marked by symbols.
suitable tools to study the processes during dehydroge-
nation and rehydrogenation reactions of NaAlH4.
2. While in most cases the results obtained with both
analytical techniques applied agree perfectly, in some
cases discrepancies were observed which can be
rationalized by taking the peculiarities of the individual
techniques into account. The skin effect occurring in
larger metallic particles in the NMR experiments and
the fact that amorphous phases are not considered in the
quantitative Rietveld refinement lead to apparently
Fig. 8. Rietveld plot after final refinement of the sample B6. Observed data (cross), calculated data (line), and the difference curve. The reflection positions
for the two phases are marked by vertical lines: upper row, Al; bottom row, NaAlH4.

´
B. Bogdanovic et al. / Journal of Alloys and Compounds 350 (2003) 246–255
254
observed in the X-ray powder diagrams to a significant
extent. In addition, non-identified phases with a content
of around 1–2 wt.% were observed. These minority
phases, however, could not clearly be associated with
particular steps in the dehydrogenation/rehydrogenation
cycle and are thus probably not crucial in understanding
the system behavior. Of particular interest, however, is
a shoulder on the right side of the main Al reflection,
the intensity of which increases with the amount of the
applied catalyst, and which can be tentatively assigned
to an Al–Ti alloy.
4. Rehydrogenation of a completely dehydrogenated sam-
ple leads to the formation of NaAlH4 and Na3AlH6
while some Al is left, but only traces of NaH are
observed. This means that the rehydrogenation of NaH
with Al to Na3AlH6 (Eq. (1), 1st step) runs almost to
completion, while the reaction of Na3AlH6, Al and
hydrogen to NaAlH4 (2nd step) stops before it is
completed. This incomplete rehydrogenation of
Na3AlH6 /Al is responsible for the storage capacity in
cycle tests being lower than theoretically expected. It is
suggested that this results from relatively large particles
with an aluminum core which is covered by NaAlH4.
The aluminum is thus effectively separated from the
other reagent, Na3AlH6, and mass transfer limitations
prevent full rehydrogenation. In agreement with this
hypothesis, the signal of Na3AlH6 almost completely
disappears from the XRD pattern and the NMR spec-
trum, when a rehydrogenated sample is ball-milled with
fresh Al powder and then hydrogenated again.
Fig. 9. 23Na MAS NMR spectra (60 scans, 20 s recycle delay) of the
rehydrogenated samples B4–B6 (Table 1) recorded at a spinning rate of
13 kHz.
contradictory results for certain states of the potential
hydrogen storage material. Both techniques should be
applied in combination to obtain an adequate picture of
the dehydrogenation and rehydrogenation processes.
3. Only NaAlH4, Na3AlH6, Al metal and NaH were
Acknowledgements
This work was partially supported by Adam Opel AG,
Bomin Solar and the Fonds der Chemischen Industrie, in
addition to the basic funding by the Max Planck
Gesellschaft. We would also like to acknowledge the input
of the other members of the hydrogen storage team, W.
Schmidt, J.M. von Colbe de Bellosta, M. Mamatha, and A.
Taguchi.
References
´
[1] B. Bogdanovic, M. Schwickardi, J. Alloys Comp. 253–254 (1997)
1.
´
´
[2] B. Bogdanovic, R.A. Brand, A. Marjanovic, M. Schwickardi, J.
¨
Tolle, J. Alloys Comp. 302 (2000) 36.
[3] C.M. Jensen, K.J. Gross, Appl. Phys. A 72 (2001) 213.
[4] K.J. Gross, G.J. Thomas, C.M. Jensen, J. Alloys Comp. 330–332
(2002) 683.
[5] G. Sandrock, K.J. Gross, G.J. Thomas, C. Jensen, D. Meeker, S.
Takara, J. Alloys Comp. 330 (2002) 696.
Fig. 10. 27Al MAS NMR spectra (2400 scans, 5 s recycle delay) of the
rehydrogenated samples B4–B6 (Table 1) recorded at a spinning rate of
11 kHz. The resonance line of the metallic aluminium in the spectrum of
sample B4 is cut at 35% of the full height.
[6] B. Bogdanovic, G. Sandrock, MRS Bulletin, September 2002 (in
´
press).
[7] K.J. Gross, S. Guthrie, S. Takara, G.J. Thomas, J. Alloys Comp. 297
(2000) 270.

´
B. Bogdanovic et al. / Journal of Alloys and Compounds 350 (2003) 246–255
255
[8] V.P. Balema, J.W. Wiench, K.W. Dennis, M. Pruski, V.K. Pecharsky,
J. Alloys Comp. 329 (2001) 108.
[13] V.P. Tarasov, G.A. Kirakosyan, Russ. J. Inorg. Chem. 42 (1997)
1223.
´
[9] J. Rodrıguez-Carvajal, in: Abstracts of the Satellite Meeting on
¨
´
´
[14] E. Ronnebro, D. Noreus, K. Kadir, A. Reiser, B. Bogdanovic, J.
Powder Diffraction of the XV Congress of the IUCr, Toulouse,
1990, p. 127.
[10] J.W. Lauher, D. Dougherty, P.J. Herley, Acta Crystallogr. B35
(1979) 1454.
[11] D. Sun, T. Kiyobayashi, H.T. Takeshita, N. Kuriyama, C.M. Jensen,
J. Alloys Comp. 337 (2002) L8.
Alloys Comp. 299 (2000) 101.
[15] C.G. Shull, E.O. Wollan, G.A. Morton, W.L. Davidson, Phys. Rev.
73 (1948) 842.
[16] L. Kellberg, H. Bildsøe, H.J. Jakobsen, J. Chem. Soc., Chem.
Commun. (1990) 1294.
[17] J. Schuster, H. Iper, Z. Metallkd. 81 (1990) 389.
[12] V.P. Tarasov, S.I. Bakum, V.I. Privalov, Yu.B. Muralev, A.A.
Samoilenko, Russ. J. Inorg. Chem. 41 (1996) 1104.