Synchrotron X-ray studies of Al1-yTiy formation and re-hydriding inhibition in Ti-enhanced NaAlH4

Article abstract of DOI:10.1021/jp051031p

NaAlH₄ samples with Ti additives (TiCl₃, TiF ₃, and Ti(OBu)₄) have been investigated by synchrotron X-ray diffraction in order to unveil the nature of Ti. No crystalline Ti-containing phases were observed after ball milling of NaAlH₄ with the additives, neither as a solid solution in NaAlH₄ nor as secondary phases. However, after cycling, a high-angle shoulder of Al is observed in the same position with 10% TiCl₃ as that with 2% Ti(OBu)₄, but with considerably higher intensity, indicating that the shoulder is caused by Ti. After prolonged reabsorption, there is only a small fraction of free Al phase left to react with Na₃AlH₆, whereas the shoulder caused by Al_1-yTi_y is dominating. The Ti-containing phase causing the shoulder therefore contains less Ti than Al₃Ti, and the aluminum in this phase is too strongly bound to react with Na ₃AlH₆ to form NaAlH₄. The composition of the Al_1-yTi_y phase is estimated from quantitative phase analysis of powder X-ray diffraction data to be Al_0.85Ti _0.15. Formation of this phase may explain the reduction of capacity beyond the theoretical reduction from the dead weight of the additive and the reaction between the additive and NaAlH₄. ? 2005 American Chemical Society.

Full text of DOI:10.1021/jp051031p

15780
J. Phys. Chem. B 2005, 109, 15780-15785
Synchrotron X-ray Studies of Al1-yTiy Formation and Re-hydriding Inhibition in
Ti-Enhanced NaAlH4
,†
†
‡
‡
Hendrik W. Brinks,* Bjørn C. Hauback, Sesha S. Srinivasan, and Craig M. Jensen
Institute for Energy Technology, P. O. Box 40, Kjeller, NO-2027 Norway, and Department of Chemistry,
UniVersity of Hawaii, Honolulu, Hawaii 96822
ReceiVed: March 1, 2005; In Final Form: May 9, 2005
NaAlH
diffraction in order to unveil the nature of Ti. No crystalline Ti-containing phases were observed after ball
milling of NaAlH with the additives, neither as a solid solution in NaAlH nor as secondary phases. However,
after cycling, a high-angle shoulder of Al is observed in the same position with 10% TiCl as that with 2%
Ti(OBu) , but with considerably higher intensity, indicating that the shoulder is caused by Ti. After prolonged
reabsorption, there is only a small fraction of free Al phase left to react with Na AlH , whereas the shoulder
caused by Al1-yTi is dominating. The Ti-containing phase causing the shoulder therefore contains less Ti
than Al Ti, and the aluminum in this phase is too strongly bound to react with Na AlH to form NaAlH . The
composition of the Al1-yTi phase is estimated from quantitative phase analysis of powder X-ray diffraction
data to be Al0.85Ti0.15. Formation of this phase may explain the reduction of capacity beyond the theoretical
4 3 3 4
samples with Ti additives (TiCl , TiF , and Ti(OBu) ) have been investigated by synchrotron X-ray
4
4
3
4
3
6
y
3
3
6
4
y
4
reduction from the dead weight of the additive and the reaction between the additive and NaAlH .
0
1
. Introduction
of TiIII/IV to Ti without changing the valence of Al.7 57Fe
M o¨ ssbauer spectroscopy investigations confirm formation of bcc
Fe from addition of both Ti(OBu)4 and Fe(OEt)2 to NaAlH4
and suggest formation of an Fe-Al alloy during dehydrogen-
The development of lightweight, high reversible capacity
hydrogen storage materials is of central importance to the
actualization of practical hydrogen-powered vehicles. In 1997,
Bogdanovic and Schwickardi reported that the elimination of
hydrogen from solid NaAlH4 is markedly accelerated and
rendered reversible under moderate conditions upon mixing the
hydride with a few mole percent of selected transition metal
complexes. Since that time, aluminum-based complex hydrides,
7
1
ation.
For samples prepared by ball milling, recent X-ray absorption
studies indicate Ti , at least after cycling of NaAlH4 with Ti
0
8-10
additives.
Thorough powder X-ray diffraction (PXD) studies
of fresh and cycled NaAlH4 samples with different additives
do not indicate any significant bulk solid solution of Ti in Na
or Al sites from precise determination of the unit-cell dimen-
Neutron diffraction data, with intensities strongly
dependent on an eventual solid solution of Ti in NaAlD4, are
in line with that. No crystalline Ti-containing phases were
detected in samples that were ball-milled but not cycled.1
However, after cycling at least part of the Ti is forming a phase
with Al after cycling with a lower Ti-content than Al3Ti.11 The
content of this Ti-Al phase has been reported to increase with
increasing Ti additive. Similar observations were made in
studies of the synthesis of TiAl3 from a suspension of TiCl3
and LiAlH4 in mesitylene heated to 550 °C. In this study, Al
was detected by PXD in the precipitate before heating, but
neither Ti nor any Ti-Al alloy were detected at this point. The
main part of the Ti compounds added to NaAlH4 is probably
“
alanates”, have been regarded as a highly promising group of
materials for onboard hydrogen-storage applications. Hydrogen
cycling capacities of 3-4 wt % have been achieved for
Ti-doped NaAlH4 at 120-160 °C with relatively good kinetics.
1
1,12
sions.
2
-5
1
1
NaAlH4 decomposes according to the three sequential steps
1,12
seen in eqs 1-3. The first two reactions, releasing in total
5
hydrogen storage at moderately low temperatures. Two-thirds
of this hydrogen is released in the first reaction.
.6 wt % hydrogen, are regarded as useable for reversible
1
3
1
2
NaAlH ) / Na AlH + / Al + H
2
(1)
(2)
(3)
4
3
3
6
3
1
4
1
1
1
/3Na AlH ) NaH + / Al + / H
3 6 3 2 2
1
NaH ) Na + / H
2
2
0
reduced to Ti in the ball-milling process and forms an Al-rich
Ti-Al alloy after the heat treatment during cycling. The fact
Ti can be introduced into the hydride either by mixing
NaAlH4 with the Ti-based additive (often titanium halides) in
a dispersion or by ball milling. Samples prepared in these ways
have been studied, and partial results on the state of Ti have
been achieved. For the former technique, hydrogen evolution
during the reaction was determined in accordance with reduction
that Ti-Al alloys can be prepared by ball milling of the
1
6
15-21
elements
supports this hypothesis. However, still the
mechanism of the enhancement of the kinetics is not revealed.
To further unveil the nature of the titanium species that arises
from ball milling NaAlH with titanium additives and perhaps
4
hence extend the knowledge of the mechanism of the action of
*
To whom correspondence should be addressed. Telephone:
the additives, cycled samples with TiCl3, TiF3, and Ti(OBu)4
additives were analyzed in detail by synchrotron radiation PXD
(SR-PXD).
+
47 63 80 64 99. Fax: +47 63 81 09 20. E-mail: hwbrinks@ife.no.
†
Institute for Energy Technology.
University of Hawaii.
‡
1
0.1021/jp051031p CCC: $30.25 © 2005 American Chemical Society
Published on Web 07/30/2005

Ti-Enhanced NaAlH4
J. Phys. Chem. B, Vol. 109, No. 33, 2005 15781
TABLE 1: Mole Fractions (%) of the Phases Based on Rietveld Refinements
sample no.
additive
no.of cycles
% NaAlH
4
% Na
3
AlH
6
% Al/Al1-yTi
y
% NaCl
1
2
3
10% TiCl
10% TiCl
3
0
3
0
58.9
14.4
59.0
71.9
22.4
46.8
36.6
19.8
18.7
24.5
3
14.2
4.4
5.9
a
10% TiF
2% Ti(OBu)
3
4
4
101
2.0
a
NaF was not observed.
earlier observations11 and indicates that there is no significant
solid solution of Ti in the alanate. In addition, no Ti-containing
phases were observed in the uncycled samples. Hence, no
conclusions can be drawn from SR-PXD on the state of Ti
directly after ball milling. The fit of the Rietveld refinement of
the uncycled sample 1 is shown in Figure 1.
The addition of TiCl3 results in formation of crystalline Al
and NaCl. The SR-PXD reflections from NaCl are very broad
after ball milling, but after cycling they are significantly sharper;
cf. Figure 2. On the basis of microstructural analysis of the
SR-PXD data, the broadening was found to originate both from
strain and small crystallite size. The crystallite size was
estimated to increase from 11 nm after ball milling until
TABLE 2: Unit-Cell Dimensions Determined by Rietveld
Refinements of Synchrotron Radiation PXD Data of the
Ti-Enhanced NaAlH
NaAlH
a (Å)
4
Samples
4
Na
V (Å3)
3
AlH
6
Al
Al1-yTi
a (Å)
y
sample no.
c (Å)
a (Å)
1
2
3
4
5.0237(1) 11.3507(2)
5.0234(1) 11.3514(2) 232.34(1) 4.0493(1) 4.0368(1)
5.0235(1) 11.3500(2) 232.50(2) 4.0493(1)
4.0489(1)
5.0237(1) 11.3503(1) 232.39(1) 4.0490(-) 4.0344(4)
a
Estimated standard deviations are given in parentheses. For
comparison, the values for the pure phases determined at the same
instrument are as follows: a ) 5.0232(1) Å and c ) 11.3483(1) Å for
NaAlH
4
3 6
, V ) 232.50(2) Å for Na AlH
, and a ) 4.0490 Å for Al.11
2
70 nm after 3 cycles. Hence, the heat treatment in the cycling
2
. Experimental Section
process results in a considerable increased crystallite size for
NaCl. Al is formed in the same reaction, but the reflections are
much sharper and the change in half-width of the reflections
during cycling is less pronounced. The crystallite sizes of Al
and NaAlH4 were estimated to 56 and 170 nm after ball milling,
and the sizes increase during cycling. More details on the
microstructure of Ti-enhanced NaAlH4 will be given in a
forthcoming paper.
All reactions and operations were performed under argon in
a glovebox or using standard Schlenk techniques with oxygen
and water free solvents. NaAlH4 was purchased from Albemarle
Corp. and recrystallized with THF/pentane before use. TiCl3
and TiF3 were purchased from Aldrich Chemicals Inc. and used
without further purification. TiCl3 and TiF3 were added to
NaAlH4 by ball milling 30 and 60 min, respectively, in a Fritsch
pulverizette 7 at 350 rpm. The ball to sample mass ratio was
approximately 20:1.
The reaction can be assumed to be
The cycled samples with TiCl3 were dehydrided at 160 °C
against 1 bar pressure for 3 h and hydrided at 115 bar at
NaAlH + 0.1TiCl ) 0.7NaAlH + 0.3NaCl + 0.3Al +
4
3
4
0
.1“Ti”(am) + 0.6H2(g) (4)
120 °C for 12 h. The NaAlH4 with Ti(OBu)4 additive was cycled
1
00 cycles with the same conditions (same sample as in ref 3),
After normalization, the expected mole percent of the crystalline
phases for this reaction is 53.8% NaAlH4, 23.1% NaCl, and
and the last rehydrogenation (cycle 101) was carried out at 85
bar and 120 °C for 65 h.
2
3.1% Al. The refined values based on the SR-PXD data for
SR-PXD data at 22 °C were collected at the Swiss-
Norwegian beam line (station BM1B) at the European Syn-
chrotron Radiation Facility (ESRF) in Grenoble, France. The
samples were kept in rotating 0.5 mm boron-silica-glass
capillaries. Intensities were measured in steps of ∆(2θ) ) 0.004
and 0.005°. The wavelength of 0.49956 Å was obtained from a
channel-cut Si(111) monochromator.
sample 1 are 58.9, 18.7, and 22.4%, respectively. This indicates
that eq 4 is a good approximation of the reaction when NaAlH4
is ball-milled with 10% TiCl3.
For the TiF3-enhanced NaAlH4 (sample 3), crystalline NaF
is not formed as expected by analogy to eq 4. This was earlier
11
observed for 6% TiF3, and even 10% is not enough to observe
any NaF. The reason for this may be that the crystallite size is
even smaller than for NaCl and it becomes below the detection
limit of SR-PXD. The background is slightly elevated around
the expected positions for the NaF reflections. Furthermore,
preliminary in-situ SR-PXD experiments indicate that NaF is
formed upon heating in a vacuum and that the reflections
gradually sharpen by heating.
Rietveld refinements were carried out using the program
22
Fullprof (version 2.80). X-ray form factors were taken from
the Fullprof library. Thompson-Cox-Hastings pseudo-Voigt
profile functions were used, and backgrounds were modeled
by interpolation between manually chosen points. Information
of the microstructure was obtained from isotropic broadening
effects included in the Rietveld refinements. The instrumental
resolution was determined with a LaB6 standard.
Small amounts of Na3AlH6 were observed after ball milling
with TiF3 (sample 3). By assuming the reaction
3
. Results
NaAlH + 0.1TiF ) 0.7NaAlH + 0.3Al +
4
3
4
The samples that were analyzed by SR-PXD were NaAlH4
0
.3“NaF”(am) + 0.1“Ti”(am) + 0.6 H2(g) (5)
with 10 mol % TiCl3, uncycled (sample 1) and after 3 cycles
sample 2), 10 mol % TiF3 uncycled (sample 3) and 2 mol %
(
Ti(OBu)4 (sample 4) after 101 cycles. The mole fractions of
each phase present in the samples and selected unit-cell
dimensions from the Rietveld refinements are shown in Tables
and furthermore allowing partial thermal decomposition during
ball milling according to eq 1, the amounts of the resulting
phases can be estimated. Rietveld refinements of the SR-PXD
data gave 4.4 mol % Na3AlH6. From eqs 5 and 1 this
corresponds to 56.8 and 38.8 mol % of NaAlH4 and Al,
respectively. This fits very well with the observed values of
1
and 2, respectively.
The unit-cell dimensions of the samples do not deviate from
the pure samples of NaAlH4 and Na3AlH6. This is in line with

15782 J. Phys. Chem. B, Vol. 109, No. 33, 2005
Brinks et al.
Figure 1. Observed intensities (circles) and calculated intensities from
Rietveld refinements (upper line) of NaAlH with 10 mol % TiCl added
ball milling, measured at 22 °C at BM1B, ESRF. Positions of Bragg
reflections are shown with bars for NaAlH , Al, and NaCl (from top).
The difference between observed and calculated intensity are shown
with the bottom line.
Figure 3. Observed intensities (circles) and calculated intensities from
4
3
Rietveld refinements (upper line) of NaAlH with 10 mol % TiCl after
4
3
3 cycles, measured at 22 °C at BM1B, ESRF. Positions of Bragg
4
reflections are shown with bars for NaAlH4, Al, Al1-yTi , Na AlH ,
y
3
6
and NaCl (from top). The differences between observed and calculated
intensities are shown with the bottom line.
reason in this case; cf. Figure 3. All Al reflections are split in
two-a low-intensity reflection at low angles and a reflection
with a significantly higher intensity at higher angles. In the
earlier study with only 2 mol % Ti(OBu)4, there was a shoulder
at the high-angle side that was interpreted as Al1-yTiy. On the
basis of the unit-cell volume, y was estimated to 0.07.11
The low-angle reflections at 12.27 and 14.17° are in line with
pure Al (a ) 4.0490 Å). Only a small amount of pure Al is left
in sample 2. If the unlikely options of amorphous or evaporated
aluminum are rejected, the only reasonable possibility left is
that most of the Al is chemically bonded in the phase which
gives reflections at the high angle side of pure Al. Unless there
is an oxygen impurity, a reaction with Ti is most likely. The
extra peaks are also observed with two different additives and
with increasing intensities with an increasing amount of additive.
This phase has a slightly lower unit-cell dimension than Al
(
4.0368 compared to 4.0490 Å) and has the same relative
Figure 2. Comparison of the peak shape of NaCl in NaAlH
4
with
intensities as fcc Al. Al with small substitutions of Ti results in
a smaller unit-cell dimension, and the unit-cell dimensions of
1
0 mol % TiCl
3
after ball milling and after cycling.
23
both the stable D022 and the metastable L12 modification of
1
8
5
9.0 (NaAlH4) and 36.6 mol % (Al). Hence, the “crystalline
Al3Ti are smaller. These are indications of an Al-rich Al-Ti
intermetallic phase in this case. The unit-cell volume per atom
increases at the Ti-rich side of Al3Ti to a higher value than
part” of eq 5 is confirmed, and the consumption of TiF3 is fairly
complete.
1
8
Quantitative phase analysis of the cycled sample of NaAlH4
with 10 mol % TiCl3 (sample 2), which was rehydrided in the
final cycle, indicates equal amounts of NaAlH4 and Na3AlH6;
cf. Table 1. On the basis of eqs 4 and 1, the phase fractions can
be estimated to be 13.5 (NaAlH4), 13.5 (Na3AlH6), 50.0 (Al),
and 23.0% (NaCl). These are close to the observed values: 14.4,
pure Al, but if the present phase had a lower Al content than
Al3Ti, then there would have been free Al both from the ball-
milling reaction (eq 4) and from the fraction of Al after eq 4
which still has not reacted with Na3AlH6. This is not observed.
Hence, the data from a sample with high additive level proves
1
unequivocally that an Al1-yTiy phase with y < /4 is formed. If
1
4.2, 46.8, and 24.5%, respectively. After cycling, including
no Ti-Al phase were formed at all, an excess of Al should be
present at this additive level and probably a predominant amount
of Na3AlH6 would react.
rehydrogenation, only 14.4% NaAlH4 is formed compared to
the theoretical capacity of 53.9% (eq 4). This means that, even
after correction for NaCl and Al formed during ball milling,
the rehydrogenation is only 27% of the theoretical capacity of
the first step for NaAlH4/Na3AlH6 (eq 1), whereas the rehydro-
genation of the second step (eq 2; Na3AlH6/NaH) is complete,
because no NaH is observed. This interesting observation
demands further investigation.
There is certainly enough Na3AlH6 available for reaction with
Al. Generally, long diffusion paths, e.g. through a NaAlH4
product layer, may reduce the effective capacity. However, a
closer look at the Al reflections reveals that this is not the main
There is no stable phase between Al and Al3Ti according to
1
8
the phase diagram, and reports on fast quenching from melt
23
results in phase separation into Al3Ti and Al0.998Ti0.002. On
the other hand, there are several reports on attempts on
mechanical alloying in this composition range. A continuous
change in the unit-cell dimension of the more Al-rich phase
with milling time from Al to approximately the unit-cell
1
5
dimension of Al3Ti has been reported. Furthermore, the
disappearance of Ti-reflections by PXD after ball milling has
20,21
been reported.
On the basis of the quantitative phase analysis

Ti-Enhanced NaAlH4
J. Phys. Chem. B, Vol. 109, No. 33, 2005 15783
Figure 4. Mole fractions of NaAlH
different amounts of TiCl additive.
3
4
and Al1-yTi
y
based on eq 4 for
Figure 5. SR-PXD data around the strongest Al reflection for a
reabsorbed sample after 7 and 101 cycles, showing that there is less
free Al (left reflection) after 101 than after 7 cycles.
above, there is probably not much of the Al that reacts with Ti
during ball milling even though the phases are formed in the
same reaction (eq 4) and should be in close proximity in the
sample. This could be due to relatively short ball-milling times
used for reversible hydrogen storage in alanates compared to
studies on mechanical alloying of Al-Ti intermetallics. But after
a few cycles, probably because of the elevated temperature, at
least a partial reaction takes place.
tion (to be compared to eq 6) is 22.7% NaAlH4, 11.5%
Na3AlH6, and 41.3% Al1-yTiy, and the remaining is NaCl. From
Figure 4, the mole fractions of NaAlH4 and Al1-yTiy were used
separately to estimate y values of 0.153 and 0.173. A similar
curve for Na3AlH6 gave a y value of 0.138. The quantitative
phase analysis points to the existence of Al0.85Ti0.15 in the cycled
NaAlH4 sample with 10% TiCl3 additives. The unit-cell dimen-
sion of this Al “shoulder” is the same as that for the cycled
NaAlH4 sample with 2% Ti(OBu)4, and the composition is
probably similar. The present quantitative phase analysis, based
on SR-PXD, gives a more accurate composition estimate than
the rough estimation from linear unit-cell volume intrapolation
between Al and Al3Ti with the L12 structure, in particular for
samples with high amounts of additive. The asymmetric profile
of the first Al1-yTiy reflection in Figure 3 may indicate that the
composition of this phase is not completely homogeneous
throughout the sample.
Apparently, the Al1-yTiy phase is so stable that Al is trapped
and will not react with Na3AlH6. This will reduce the reversible
storage capacity beyond the reduction in capacity from the dead
weight from the additive and the reaction of the additive with
NaAlH4 during ball milling. The total reaction of the initial ball
1
1
1
milling reaction and the formation of Al1-yTiy with y < /4
during cycling could be formulated like
NaAlH + xTiCl ) (1 - kx)NaAlH + 3xNaCl +
4
3
4
(
x/y)Al1-yTi + lxNa AlH + (2k - 3l)xH (6)
y 3 6 2
NaAlH4 with 2 mol % Ti(OBu)4 (sample 4) was in the 101th
cycle re-hydrogenated at 120 °C and 86 bar pressure in a
Sieverts-type apparatus with excellent temperature stability for
with
6
5 h. Of the final absorption 84 and 92% are completed during
3
k ) (3 - 6y)/2y and l ) (1 - y)/2y - /
the first 12 and 20 h, respectively.
2
The quantitative phase analysis of this sample resulted in
71.9 mol % NaAlH4, 5.9 mol % Na3AlH6, and 19.8 mol % Al
(15.1 mol % present as Al1-yTiy). In addition the sample
included 2 mol % NaCl impurity. By assuming that the additive
destroys 8 mol % NaAlH4 and forms crystalline Al and
amorphous phases such as NaOBu and Ti, the phase fractions
can from eq 1 be estimated to 74.3, 5.9, and 19.8 mol %,
respectively. This corresponds to 81% complete re-hydrogena-
tion in step 1 of NaAlH4.
This equation is based on a complete consumption of all Ti in
order to form a homogeneous Al1-yTiy phase. The excess Al
(compared to y ) 0.25) is taken from NaAlH4, and the amount
of Na3AlH6, with Na:Al ) 3:1, is balanced to give the same
amounts of Al and Ti on both sides of the equation.
A more Al-rich phase (lower y) leads to less regenerated
NaAlH4, as shown for 2, 6, and 10% additive level in Figure 4,
and reduced capacity on the first stage of NaAlH4. These curves
may be used to estimate y based on the phase composition. The
estimated y value is then used in the Rietveld refinements to
generate new and more accurate phase compositions. By this
recursive procedure a value for y is obtained. Similarly, the
reduced capacities during cycling in Sieverts-type apparatus may
also be used to estimate y, on the assumptions that all free Al
is reacted with Na3AlH6 and that Na3AlH6 is completely
dehydrided in the experimental conditions chosen for the
cycling. However, reduced capacity due to reasons other than
Al1-yTiy formation will lead to an underestimation of y.
The present data were used for the estimation of y. After
corrections for assumed complete reaction of free Al (5.5 mol
Compared to the SR-PXD data after 7 cycles for the same
1
1
sample, the peak widths of NaAlH4 and Na3AlH6 for the
sample after 101 cycles are similar. However, the Al peaks
differ. In Figure 5, SR-PXD data around the strongest Al peak
of the 7 and 101 cycle samples are compared after background
subtraction and normalizing to the strongest peak from NaAlH4.
Because of the slightly different wavelengths of the experiments,
the data are presented on the Q-scale (2π/d). After 101 cycles
the content of free Al is clearly less than after 7 cycles, probably
because of the extended re-hydrogenation time. This brings
about better opportunities to examine the shoulder. There appear
-
1
to be two shoulders: at a ) 4.038 Å (Q ) 2.695 Å ) and at
-
1
%) with Na3AlH6 (reversed eq 1), the remaining phase composi-
a ) 4.018 Å (Q ) 2.709 Å ). Similar double shoulders are

15784 J. Phys. Chem. B, Vol. 109, No. 33, 2005
Brinks et al.
also observed at the high-angle side of the second strongest Al
reflection. Both shoulders were included in the refinements for
estimation of the fraction of Al/Al1-yTiy. A possibility is that,
by extended cycling, aluminum and titanium may continue the
reaction and several types of Al-rich intermetallics are formed.
In the dehydrided sample after 100 cycles, the mole fractions
were estimated to be 54.6, 8.7, and 35.3% Al, Na3AlH6, and
NaH, respectively.11 This corresponds to a 59% complete
desorption of Na3AlH6. Since the mol fractions after the next
absorption corresponds to 81%, the capacity could, as a first
approximation, be estimated from the capacities of the two
reaction steps to be 3.8 wt %. Corrected for the prolonged
absorption in the last reabsorption, the estimation from quantita-
tive phase analysis is close to the observed values for the last
cycles, which is approximately 3.4 wt %.
4
. Discussion
Figure 6. Theoretical hydrogen storage capacity of NaAlH
4
with
Three different X-ray absorption studies near the Ti edge
different amounts of TiCl additive as a function of the Ti content in
3
III
0
indicate reduction of Ti to Ti after the ball-milling reaction,
but the interpretation of the extended X-ray absorption fine
structures deviate considerably.8 A general drawback of this
y
Al1-yTi .
-10
0
directly after ball milling is interpreted as an amorphous
technique is that because the Ti edge is at lower energy than
2
9
Ti-Al phase of undefined composition.
the edge for higher Ti valences, a smaller amount of Ti in higher
0
valences would be masked by the Ti contribution. Nevertheless,
The present results show that, after cycling a metastable phase
with a composition between Al3Ti and Al is formed. This phase
is formed for Ti additions as high as 10 mol %. Estimations
from the quantitative phase analysis give a composition of Al0.85-
Ti0.15, and this phase is characterized by a unit-cell dimension
it is clear that most of the titanium precursor is reduced to the
metallic state by the ball milling with the reducing agent
NaAlH4.
Similar results were recently shown by pressure measurements
24
of approximately 4.0365 Å, both with the Ti(OBu) and the
during ball milling. This study showed a pressure increase in
accordance with complete reduction of TiClx to Ti. But without
PXD characterization of the product, it is still a possibility for
partial thermal decomposition to Na3AlH6 during the prolonged
milling. This would eventually affect the calculated number of
hydrogens released per Ti considerably.
4
TiCl3 additive. This phase can account for more than 80% of
the Ti introduced into the sample. Furthermore, the positions
of Al and Ti have been shown to be correlated in a cycled
NaAlH4 with TiF3 additive from energy-dispersive X-ray
analysis in a recent transmission electron microscopy study.30
III
EPR experiments, which are sensitive to Ti , show traces of
The effect of Al0.85Ti0.15 on the kinetics of the sodium alanate
system is at the moment not clear. There is always a possibility
that a minor fraction of Ti is causing the improvement, whereas
the major fraction is detected by PXD and other techniques.
Nevertheless, after cycling Ti-enhanced NaAlH4, most of the
Ti is present as a solid solution in Al, and it is a reasonable
possibility that Ti from this phase is of importance to the
enhanced kinetics. At the moment, it is not clear whether the
composition of the Al-Ti phase is constant during desorption
and absorption.
TiIII after ball milling with TiCl3 and larger amounts of Ti
III
25
III
after ball milling with TiF3. In both cases, Ti are replaced
by signals for Ti after a few cycles. This is interpreted as
0
unreacted additive that is consumed during cycling.
The present study using quantitative phase analysis of
SR-PXD data shows that ball milling of NaAlH4 with TiCl3
and TiF3 leads to close to complete reaction of the Ti precursors
with NaAlH4, such that NaCl and Al are formed. Neither Na
III
nor Al are found to be replaced by Ti. This implies that Ti is
reduced, either partially to e.g. TiHx or Ti1-yAlyHx or completely
to Ti (Ti or a Ti-rich Ti-Al alloy). Because no crystalline
Ti-containing phases are found, it is not possible to conclude
from SR-PXD the state of Ti directly after ball milling with
NaAlH4.
When Al1-yTiy is formed during cycling, this reduces the
storage capacity of the first step of NaAlH4 because of
insufficient Al for the reaction with Na3AlH6. Combined with
the capacity loss from the dead weight of the additive and
reaction of the additive according to eq 4, the hydrogen storage
capacity of NaAlH4 with TiCl3 additive as a function of y in
Al1-yTiy is shown in Figure 6. A decreasing y leads to a
considerable loss of capacity; e.g. formation of Al0.85Ti0.15 in a
0
Recent experiments with 0.915 Ti + 0.085 Al in methanol
indicate removal of pure Al within 4 h of ball milling and
formation of an amorphous phase, which may contain hydrogen,
after 12 h.26 It has been reported that stable Ti-Al phases with
sample with 2 mol % TiCl will give a maximum storage
3
27
capacity of 4.73 wt %, whereas with 10 mol % only 1.94 wt %
may be maintained.
up to 45% Al may be hydrided, and metastable phases
achieved by mechanical alloying up to 65% Al may be
hydrided.2 A considerable Al content would have caused a shift
in the Al:NaCl ratio, which in the present work is shown to be
in close agreement with eq 4 for sample 1. Similarly, a large
hydrogen content in the Ti-containing phase is expected to give
8
Certainly, additional decreased capacity due to long diffusion
paths and resulting incomplete reaction may be important.
1
3
Bogdanovic et al. did observe Na3AlH6 after absorption, and
ball milling of this sample did not remove Na3AlH6 completely,
but excess Al added by ball milling at this stage did. This could
also be interpreted on the basis of the present model, because
it is reasonable to believe that ball milling does not release free
Al from Al1-yTiy, which probably was present in the sample
0
a shift in the near-edge X-ray absorption compared to Ti . These
considerations indicate formation of amorphous titanium, pos-
sibly with small amounts of Al and H, after ball milling. In a
recent report, a wide feature in the PXD diagram measured

Ti-Enhanced NaAlH4
J. Phys. Chem. B, Vol. 109, No. 33, 2005 15785
after cycling. Introduction of free Al, however, may consume
the remaining Na3AlH6.
There are literature data available on cycling capacities for
(6) Jensen, C. M.; Zidan, R.; Mariels, N.; Hee, A.; Hagen, C. Int. J.
Hydrogen Energy 1999, 24, 461-465.
(
7) Bogdanovic, B.; Brand, R. A.; Marjanovic, A.; Schwickardi, M.;
Tolle, J. J. Alloys Compd. 2000, 302, 36-58.
8) Graetz, J.; Reilly, J. J.; Johnson, J.; Ignatov, A. Y.; Tyson, T. A.
Appl. Phys. Lett. 2004, 85, 500-502.
9) L e´ on, A.; Kircher, O.; Rothe, J.; Fichtner, M. J. Phys. Chem. B
2004, 108, 16372-16376.
10) Felderhoff, M.; Klementiev, K.; Grunert, W.; Spliethoff, B.; Tesche,
NaAlH4. Bogdanovic’s original data with 2 mol % TiCl3 give
(
1
4
.2 wt %, which indicates y ) 0.10, assuming no limitations
(
in diffusion. Sandrock et al. studied the storage capacity as a
function of additive level and found a reduced capacity which
2
(
corresponds to y ) 0.18 in the framework of eq 6. These results
are in the same range as the present estimation based on
quantitative phase analysis.
B.; von Colbe, J. M. B.; Bogdanovic, B.; Hartel, M.; Pommerin, A.; Schuth,
F.; Weidenthaler, C. Phys. Chem. Chem. Phys 2004, 6, 4369-4374.
(11) Brinks, H. W.; Jensen, C. M.; Srinivasan, S. S.; Hauback, B. C.;
Blanchard, D.; Murphy, K. J. Alloys Compd. 2004, 376, 215-221.
By ball milling, it is possible to induce a variety of chemical
reactions without reaching the thermodynamically most stable
state; e.g. ball milling of LiAlD4 and 2NaH recently has been
shown to partly react to NaAlD4 and LiH, which after heat
(
12) Weidenthaler, C.; Pommerin, A.; Felderhoff, M.; Bogdanovic, B.;
Schuth, F. Phys. Chem. Chem. Phys 2003, 5, 5149-5153.
13) Bogdanovic, B.; Felderhoff, M.; Germann, M.; Hartel, M.; Pom-
(
merin, A.; Schuth, F.; Weidenthaler, C.; Zibrowius, B. J. Alloys Compd.
2003, 350, 246-255.
3
1
treatment under pressure further reacts to Na2LiAlD4H2.
Present and earlier experimental evidence indicate that mainly
(
14) Haber, J. A.; Crane, J. L.; Buhro, W. E.; Frey, C. A.; Sastry, S. M.
L.; Balbach, J. J.; Conradi, M. S. AdV. Mater. 1996, 8, 163-166.
15) Zhang, F.; Lu, L.; Lai, M. O. J. Alloys Compd. 2000, 297, 211-
218.
(16) Klassen, T.; Oehring, M.; Bormann, R. Acta Mater. 1997, 45,
935-3948.
0
0
a redox reaction to Al , Ti , and NaCl happens during ball
milling of NaAlH4 and TiCl3. This reaction is subsequently,
during the heat treatment in the cycling, followed by another
thermodynamically favorable (see e.g. ref 18) reaction: from
Al and Ti to an Al-rich Al1-yTiy.
(
3
5
(
2.
17) Klassen, T.; Oehring, M.; Bormann, R. J. Mater. Res. 1994, 9, 47-
Metastable Al-Ti phases are relatively stable at elevated
temperature after mechanical alloying, but there is a possibility
that cycling at a different temperature could give a different
composition of Al1-yTiy and hence different capacity than
achieved at 160 °C in the present report. The capacity of NaAlH4
(
18) Oehring, M.; Klassen, T.; Bormann, R. J. Mater. Res. 1993, 8,
2819-2829.
(19) Oehring, M.; Yan, Z. H.; Klassen, T.; Bormann, R. Phys. Stat. Solidi
A 1992, 131, 671-689.
(20) Choi, J. H.; Moon, K. I.; Kim, J. K.; Oh, Y. M.; Suh, J. H.; Kim,
S. J. J. Alloys Compd. 2001, 315, 178-186.
(21) Fan, G. J.; Gao, W. N.; Quan, M. X.; Hu, Z. Q. Mater. Lett. 1995,
23, 33-37.
+
2 mol % Ti(OBu)4 has, though, been shown to be quite stable
at 3.5-4.0 wt % during many cycles.
3
(
(
(
22) Rodr ´ı guez-Carvajal, J. Physica B 1993, 192, 55-69.
23) Tonejc, A.; Bonefacic, A. Scr. Metall. 1969, 3, 145-147.
24) Bellosta von Colbe, J. M.; Bogdanovic, B.; Felderhoff, M.;
Acknowledgment. The skilful assistance from the project
team at the Swiss-Norwegian Beam Line, ESRF, is gratefully
acknowledged. S.S.S. and C.M.J. gratefully acknowledge fi-
nancial support received from the Office of Hydrogen, Fuel Cells
and Infrastructure Technologies of the U.S. Department of
Energy.
Pommerin, A.; Schuth, F. J. Alloys Compd. 2004, 370, 104-109.
25) Jensen, C. M.; Sun, D.; Srinivasan, S. S.; Wang, P.; Murphy, K.;
(
Wang, Z.; Eberhard, M.; Naghipour, A. Unpublished results.
(26) Morales-Hern a´ ndez, J.; Vel a´ zquez-Salazar, J.; Garc ´ı a-Gonz a´ lez, L.;
Espinoza-Beltr a´ n, F. J.; Berceinas-S a´ nchez, J. D. O.; Munoz-Saldana, J.
J. Alloys Compd. 2005, 388, 266-273.
(
79.
27) Hashi, K.; Ishikawa, K.; Aoki, K. Met. Mater. Korea 2001, 7, 175-
References and Notes
1
(
28) Hashi, K.; Ishikawa, K.; Suzuki, K.; Aoki, K. Mater. Trans. 2002,
(
(
1) Bogdanovic, B.; Schwickardi, M. J. Alloys Compd. 1997, 253, 1-9.
2) Sandrock, G.; Gross, K.; Thomas, G. J. Alloys Compd. 2002, 339,
4
3, 2734-2740.
(29) Haiduc, A. G.; Stil, H. A.; Schwarz, M. A.; Paulus, P.; Geerlings,
2
99-308.
3) Srinivasan, S. S.; Brinks, H. W.; Hauback, B. C.; Sun, D.; Jensen,
C. M. J. Alloys Compd. 2004, 377, 283-289.
4) Fichtner, M.; Fuhr, O.; Kircher, O.; Rothe, J. Nanotechnology 2003,
J. J. C. J. Alloys Compd. 2005, 393, 252-263.
(
(30) Andrei, C. M.; Walmsley, J. C.; Brinks, H. W.; Holmestad, R.;
Srinivasan, S. S.; Jensen, C. M.; Hauback, B. C. Appl. Phys. A: Mater.
Sci. Process. 2005, 80, 709-715.
(31) Brinks, H. W.; Hauback, B. C.; Jensen, C. M.; Zidan, R. J. Alloys
Compd. 2005, 392, 27-30.
(
1
4, 778-785.
(
5) Wang, P.; Jensen, C. M. J. Phys. Chem. B 2004, 108, 15827-
1
5829.