Integrated experimental-theoretical investigation of the Na-Li-Al-H system

Article abstract of DOI:10.1021/ic062032e

First-principles modeling, experimental, and thermodynamic methodologies were integrated to facilitate a fundamentally guided investigation of quaternary complex hydride compounds within the bialkali Na-Li-Al-H hydrogen storage system. The integrated approach has broad utility for the discovery, understanding, and optimization of solid-state chemical systems. Density functional theory ground-state minimizations, low-temperature powder neutron diffraction, and low-temperature synchrotron X-ray diffraction were coupled to refine the crystallographic structures for various low-temperature distorted Na₂LiAlH₆ allotropes. Direct method lattice dynamics were used to identify a stable Na₂-LiAlH₆ allotrope for thermodynamic property predictions. The results were interpreted to propose transformation pathways between this allotrope and the less stable cubic allotrope observed at room temperature. The calculated bialkali dissociation pressure relationships were compared with those determined from pressure-composition-isotherm experiments to validate the predicted thermodynamic properties. These predictions enabled computational thermodynamic modeling of Na₂LiAlH₆ and competing lower order phases within the Na-Li-Al-H system over a wide of temperature and pressure conditions. The predictions were substantiated by experimental observations of varying Na₂LiAlH₆ dehydrogenation behavior with temperature. The modeling was used to identify the most favorable reaction pathways and equilibrium products for H discharge/recharge in the Na-Li-Al-H system, and to design conditions that maximize the theoretical hydrogen reversibility within the Na-Li-Al-H system.

Full text of DOI:10.1021/ic062032e

Inorg. Chem. 2007, 46, 1401−1409
Integrated Experimental
System
−Theoretical Investigation of the Na−Li−Al−H
Susanne M. Opalka,*,† Ole M. Løvvik,‡,§ Hendrik W. Brinks,§ Paul W. Saxe,£ and Bjørn C. Hauback§
United Technologies Research Center, East Hartford, Connecticut 06108, Centre for Materials
Science and Nanotechnology, UniVersity of Oslo, N-0318 Oslo, Norway, Institute for Energy
Technology, P.O. Box 24 N-2027 Kjeller, Norway, and Materials Design, Inc., Angel Fire,
New Mexico 87710
Received October 23, 2006
First-principles modeling, experimental, and thermodynamic methodologies were integrated to facilitate a fundamentally
guided investigation of quaternary complex hydride compounds within the bialkali Na Li Al H hydrogen storage
−
− −
system. The integrated approach has broad utility for the discovery, understanding, and optimization of solid-state
chemical systems. Density functional theory ground-state minimizations, low-temperature powder neutron diffraction,
and low-temperature synchrotron X-ray diffraction were coupled to refine the crystallographic structures for various
low-temperature distorted Na2LiAlH6 allotropes. Direct method lattice dynamics were used to identify a stable Na2-
LiAlH6 allotrope for thermodynamic property predictions. The results were interpreted to propose transformation
pathways between this allotrope and the less stable cubic allotrope observed at room temperature. The calculated
bialkali dissociation pressure relationships were compared with those determined from pressure-composition-isotherm
experiments to validate the predicted thermodynamic properties. These predictions enabled computational
thermodynamic modeling of Na2LiAlH6 and competing lower order phases within the Na−Li−Al−H system over a
wide of temperature and pressure conditions. The predictions were substantiated by experimental observations of
varying Na2LiAlH6 dehydrogenation behavior with temperature. The modeling was used to identify the most favorable
reaction pathways and equilibrium products for H discharge/recharge in the Na
conditions that maximize the theoretical hydrogen reversibility within the Na Li
−
Li
−
−
Al
−
H system, and to design
−
−Al
H system.
Introduction
Na-Li-Al-H system was the prospect of developing a
mixed bialkali alanate phase that leveraged both the revers-
ibility of the sodium alanates and the higher capacity of the
lithium alanates.4-10 This rationale is evident from a van’t
Hoff plot of the equilibrium sodium and lithium alanate
dissociation pressures constructed from recent thermody-
This study will illustrate the utility of combining first-
principles, experimental, and thermodynamic methodologies
to tune the reversible hydrogen capacity within the Na-Li-
Al-H compositional system. This integration of methodolo-
gies was inspired by the need to develop high capacity
hydrogen storage media; however, the approach described
herein is broadly applicable toward the discovery of new
solid-state compounds and the investigation of their behavior.
A quaternary phase within this system, Na2LiAlH6, has been
the subject of a number of recent studies in the hydrogen
storage community.1-6 A major motivation for studies of the
(2) Hout, J.; Boily, S.; Guther, V.; Schulz, R. J. Alloys Comp. 1999, 283,
304-306.
(3) Brinks, H. W.; Hauback, B. C.; Jensen, C. M.; Zidan, R. J. Alloys
Comp. 2005, 392, 27-30.
(4) Fossdal, A.; Brinks, H. W.; Fonneløp, J. E.; Hauback, B. C. J. Alloys
Comp. 2005, 397, 135-139.
(5) Graetz, J.; Lee, Y.; Reilly, J. J.; Park, S.; Vogt, T. Phys. ReV. B 2005,
71, 1844115/1-7.
(6) Nakamura, Y.; Fossdal, A.; Brinks, H. W.; Hauback, B. C. J. Alloys
Comp. 2006, 416, 274-278.
* E-mail: opalkasm@utrc.utc.com.
† United Technologies Research Center.
‡ University of Oslo.
(7) Arroyo y de Dompablo, M. E.; Ceder, G. J. Alloys Comp. 2004, 364,
6-12.
(8) Løvvik, O. M.; Swang, O. Europhys. Lett. 2004, 67, 607-613.
(9) Løvvik, O. M.; Swang, O. J. Alloys Comp. 2005, 404-406, 757-
761.
§ Institute for Energy Technology.
£ Materials Design, Inc.
(1) Bogdanovic, B.; Schwickardi, M. J. Alloys Comp. 1997, 253-254,
(10) Løvvik, O. M.; Swang, O.; Opalka, S. M. J. Mater. Res. 2005, 20,
3199-3213.
1-9.
10.1021/ic062032e CCC: $37.00
Published on Web 01/20/2007
© 2007 American Chemical Society
Inorganic Chemistry, Vol. 46, No. 4, 2007 1401

Opalka et al.
Methods
Experimental Synthesis and Characterization. Na2LiAlD6 was
synthesized by ball milling LiAlD4 and 2NaAlD4 in a Fritsch P7
planetary ball mill with a ball to sample mass ratio of 20 and at
700 rpm for 3 h, followed by heat treatment at 180 °C and 30 bar.
All operations were performed under argon in a glove box with
less than 1 ppm O2 and H2O.
Synchrotron Radiation Powder X-Ray (SR-PXD) data at 110 K
were collected at the Swiss-Norwegian beam line (station BM1B)
at the European Synchrotron 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) )
0.003°. The wavelength 0.49957 Å was obtained from a channel-
cut Si(111) monochromator.
Figure 1. The van’t Hoff plot predicted for the first and second dissociation
reactions of lithium and sodium alanate. The wt. % H2 discharged and the
dissociation enthalpy, ∆Hdis, are listed for each reaction (from refs 11 and
12).
Powder neutron diffraction (PND) data at 8 K were collected
with the PUS instrument at the JEEPII reactor at Kjeller, Norway.
Neutrons with wavelength 1.5554 Å were obtained from a Ge-
(511) focusing monochromator. The detector unit consists of two
banks of seven position-sensitive 3He detectors, each covering 20°
in 2θ (binned in steps of 0.05°). Data were collected from 10 to
130° in 2θ. The sample was placed in a rotating cylindrical
vanadium sample holder with 5 mm diameter.
Rietveld refinements were carried out using the program Fullprof
(version 2.80).15 X-ray form factors and neutron scattering lengths
were taken from the Fullprof library. Pseudo-Voigt profile functions
were used, and the backgrounds were modeled by interpolation
between manually chosen points.
First-Principles Calculations. Two methodologies were linked
for the first-principles prediction of finite temperature thermody-
namic properties of the solid-state phases in the Na-Li-Al-H
system. These methodologies were detailed in preceding papers on
the lower order phases within the Li-Al-H ternary system,11 the
Na-H binary system,16 and the Na-Al-H ternary system.12,17 The
first methodology employed was a full structure minimization at
the ground state using the density functional theory (DFT)
implemented in the Vienna Ab initio Simulation Package (VASP)
version 4.6 with plane wave basis sets.18-20 The second methodol-
ogy generated finite temperature thermodynamic predictions of the
minimized structure using the direct method lattice dynamics
approach of Parlinski et al.,21 implemented in the Materials Design
Phonon direct method module interfaced with the VASP code22
(herein referred to as the “Phonon direct method”). This study
focused exclusively on the Na2LiAlH6 quaternary phase, which was
predicted with geometry optimizations to be the only stable
quaternary phase to exist within the Na-Li-Al-H system.8-10 The
thermodynamic predictions of lower order stable solid-state com-
pound phases with the Na-Li-Al-H systems NaAlH4, Na3AlH6,
LiAlH4, Li3AlH6, NaH, and LiH have already been reported.11,12,16
The first-principles thermodynamic property predictions of these
phases will be employed to make a comprehensive survey of
possible reactions and competitive phase stability within the Na-
Li-Al-H system.
namic predictions11,12 in Figure 1. The practical operating
1-100 bar pressure range and 25-100 °C temperature range
for hydrogen storage reversible dehydrogenation and rehy-
drogenation on-board of a transportation vehicle was used
to define an “in situ window of reversibility” in the plot. In
the first reaction of the sodium alanate system, sodium
aluminum tetrahydride, NaAlH4, dissociates to release 3.7
wt. % molecular hydrogen, H2. The predicted equilibrium
dissociation pressure of this reaction traverses the in situ
window of reversibility. However, the predicted equilibrium
pressure for second sodium alanate dissociation reaction
skims the lower left corner of the window of reversibility.
Here, the more stable trisodium aluminum hexahydride, Na3-
AlH6, dissociates to release an additional 1.9 wt. % H2. Since
both reactions are only partially reversible in Ti-activated
sodium alanate in practice,12,13 the sodium alanate system
does not meet the DOE 2006 hydrogen storage gravimetric
target of 4.5 wt. % H2. The lithium alanate system has a
greater H2 gravimetric capacity. In the first and second
dissociation reactions given in Figure 1, lithium aluminum
tetrahydride, LiAlH4, and trilithium aluminum hexahydride,
Li3AlH6, dissociate to release 5.3 and 2.6 wt. % H2,
respectively. However, as shown in Figure 1, the predicted
dissociation pressures of these less stable reactions lie
inaccessibly above the in situ window of reversibility. Recent
modeling investigations exploring the possible increase in
gravimetric capacity of the sodium alanates through lithium
substitution on the sodium sublattice, or the possible lowering
of the lithium alanates’ dissociation pressure through sodium
substitution on the lithium sublattice, predicted that Na2-
LiAlH6 is the sole stable phase in the Na-Li-Al-H
quaternary system.8-10 On the basis of these findings,
integrated theoretical-experimental methodologies were used
in this study to design the maximum theoretical hydrogen
storage capacity that could be achieved with this Na2LiAlH6
phase within the Na-Li-Al-H system.
(15) Rodr´ıguez-Carvajal, J. Phys. B 1993, 192, 55-69.
(16) Qiu, C.; Opalka, S. M.; Olson, G. B.; Anton, D. L. Int. J. Mat. Res.
2006, 97, 845-853.
(17) Opalka, S. M.; Anton, D. L. J. Alloys Comp. 2003, 356-357, 486-
489.
(18) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 47, 558-561.
(19) Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6 (1), 15-50.
(20) Kresse, G.; Furthmuller, J. Phys. ReV. B 1996, 54, 11169-11186.
(21) Parlinski, K.; Li, Z. Q.; Kawazoe, Y. Phys. ReV. Lett. 1997, 78, 4063-
4066.
(22) Parlinski, K. MedeA-Phonon Version 1.0 using Phonon Software
3.11; Materials Design, Inc.: Angel Fire, NM.
(11) Løvvik, O. M.; Opalka, S. M.; Brinks, H. W.; Hauback, B. C. Phys.
ReV. B 2004, 69, 134117/1-9.
(12) Qiu, C.; Opalka, S. M.; Olson, G. B.; Anton, D. L. Int. J. Mat. Res.
2006, 97, 1484-1494.
(13) Brinks, H. W.; Hauback, B. C.; Srinivasan, S. S.; Jensen, C. M. J.
Phys. Chem. B 2005, 109, 15780-15785.
(14) Tang, X.; Mosher, D. A.; Anton, D. L. In Materials and Technology
for Hydrogen Storage and Generation; Nazri, G-A., Ping, C., Young,
R. C., Nazri, M., Wang, J., Eds: Mater. Res. Soc. Symp. Proc. 884E;
Materials Research Society: Warrendale, PA, 2005; p GG4.4.
1402 Inorganic Chemistry, Vol. 46, No. 4, 2007

The Na-Li-Al-H System
This study sought to identify the most stable ground state Na2-
LiAlH6 allotrope to input into the Phonon direct method for reliable
thermodynamic predictions. The search for the most stable structure
followed a structural analog protocol similar to that described in
our paper on the Li-Al-H system11 and in subsequent studies on
the bialkali system.8-10 Here, Na2LiAlH6 candidate structures were
derived by substitution of ions into stoichiometrically equivalent
analogous ternary and quaternary compounds having structures
representing the widest possible range of space groups. The VASP
minimizations employed the generalized gradient approximation
(GGA) of Perdew and Wang,23 and the valence electrons were
explicitly represented with projector augmented wave (PAW)
potentials,24 using the valence configurations: Na 2s22p63s1, Li 2s1,
Al 3s23p1 and H 1s1. The plane wave cutoff energy of 780 eV and
the Gaussian smearing with an energy broadening of 0.2 eV was
used. Odd-sized k-point meshes were created by a Monkhorst-Pack
scheme for minimization of the primitive cells,25 using a k-mesh
spacing of 0.2 Å-1 or smaller. The overall total energy convergence
error for these parameters was well below 0.01 meV/unit cell. The
convergence criterion for electronic self-consistent field convergence
was 10-7 eV. A full minimization method was employed to allow
simultaneous relaxation of ionic coordinates, unit cell size, and
shape. The ground state structure was determined by minimizing
the Hellmann-Feynman forces with the conjugate gradient algo-
rithm, until all of the ionic forces were less than 0.005 eV/Å.
Following convergence, the minimizations were restarted, as
necessary, until no further ionic relaxation took place. A final total
energy calculation with accurate precision was made using the
tetrahedron smearing method with Bloechl corrections.26
The Phonon direct method was employed to predict the lattice
dynamics using the harmonic approximation on the VASP mini-
mized Na2LiAlH6 structures that had the combination of the lowest
ground state energy and the highest space group symmetry.
Displacements were made on 1 × 1 × 1 supercells for the Fm3hm
and P42/n Na2LiAlH6 structure (40 atoms each), and on a 2 × 2 ×
1 supercell of P21/n Na2LiAlH6 structure (80 atoms). A 0.1 strength
of condition factor was applied for translational invariance. Fol-
lowing the methodology described in the preceding Na-H paper,16
the integrated predicted phonon density of states was analyzed with
statistical mechanical functions to determine the vibrational con-
tribution as a function of temperature. The vibrational contribution
was added to the ground state electronic energy, to approximate
the Gibbs free energy as a function of temperature, G(T). The finite
temperature free energy G(T) values referred to the constitutive
stable element 298 K enthalpy values, are called GHSER values.16
Predicted hydride GHSER values were used in combination with
established values for elemental and gaseous phases27,28 in the
Thermo-Calc program29 to calculate thermodynamic reaction ener-
gies, equilibria, and phase diagrams.
namic description of the Na2LiAlH6 quaternary phase and
the related lower order competing or dissociation phases:
NaAlH4, Na3AlH6, NaH, LiH, etc. The recently reported
dissociation pressure data4,5 can be used to derive a partial
thermodynamic description for the Na2LiAlH6 phase, pro-
vided that the thermodynamic properties of the other phases
in the dissociation reaction are already thermodynamically
parametrized. However, the measured or predicted heats of
formation and heat capacity properties required for a
complete thermodynamic description of the Na2LiAlH6 phase
have not yet been reported. For this reason, the Na2LiAlH6
thermodynamic properties were predicted in this study using
the Phonon direct method. It was necessary to identify a
stable Na2LiAlH6 ground state structure for reliable thermo-
dynamic property predictions. This is because the contribu-
tions of the imaginary phonon frequencies generated for
metastable ground state phase structures cannot be accounted
for, with the present thermodynamic prediction methodology.
Previous refinements of near room-temperature crystal-
lographic analyses of Na2LiAlH6/Na2LiAlD6 using powder
X-ray diffraction (PXD),2,30 SR-PXD,3-5 and PND3 con-
sistently yielded a Fm3hm ordered perovskite cubic structure
with a lattice constant ranging from 7.405 to 7.407 Å. Small
imperfections in the SR-PXD data of the cubic (400)
reflection may indicate a possible small deviation from cubic
symmetry.3 However, PND data did not give any such
indication, and input of lower symmetry structures did not
improve upon the quality of the excellent refinement fit made
with Fm3hm structure.3
Density functional theory (DFT) simulations did not
corroborate the existence of a stable Fm3hm Na2LiAlH6
structure at the ground state. Earlier DFT minimization
surveys were made using analog structures representing six
different crystallographic types to predict stable global
minimum structures for bialkali aluminum hexahydride
phases.8-10 These studies were not able to distinguish
between the monoclinic P21/c and the cubic Fm3hm structures
for the Na2LiAlH6 phase,8 since the differences in cell
parameters and atomic coordinates between the two structures
were below the calculation accuracy. However, since the
monoclinic structure was slightly more stable than the cubic
one, the monoclinic structure was presented in more detail;
this also illustrated how the difference between the structures
mainly lay in the tilting of the AlH6 octahedra.
In order to probe the possible existence of lower symmetry
Na2LiAlH6 structures at temperatures approaching the ground
state, additional PND analyses were made at 8 K with the
same Na2LiAlD6 sample characterized in the recent PND
study.3 The 8 K PND data is shown with the previous 295 K
data in Figure 2. The 8 K PND data clearly indicated a
reduction of the lattice symmetry, evidenced by the extra
reflections and altered intensities of the reflections present
in the cubic structure. However, the reflections of the Fm3hm
cubic structure could not be detected to be split. The
previously refined Na2LiAlD6 structure consists of corner-
Results and Discussion
A thorough evaluation of the mixed sodium-lithium alanate
system for hydrogen storage media required a thermody-
(23) Perdew, J. P.; Chevary, J. A.; Vosko, S. H.; Jackson, K. A.; Pederson,
M. R.; Singh, D. J.; Fiolhais, C. Phys. ReV. B 1992, 46, 6671-6687.
(24) Kresse, G.; Joubert, D. Phys. ReV. B 1999, 59, 1758-1775.
(25) Monkhorst, H. J.; Pack, J. D. Phys. ReV. B 1976, 13, 5188-5192.
(26) Bloechl, P. E.; Jepsen, O.; Andersen, O. K. Phys. ReV. B 1994, 49,
16223-16233.
(27) Ansara, I.; Sundman, B. In Computer Handling and Dissemination of
Data; Glaeser, P. S., Ed.; North-Holland: Amsterdam, 1987; Vol. 1.
(28) Chase, M. W. NIST-JANAF Thermochemical Tables, 4th ed.; J. Phys.
Chem. Ref. Data, Monograph 9; American Chemical Society and
American Institute of Physics: Washington, DC, 1998; pp 1-1951.
(29) Thermo-Calc software, AB. Version Q, Stockholm, Sweden.
(30) Claudy, P.; Bonnetot, B.; Bastide, J. P.; Le´toffe´, J. M. Mat. Res. Bull.
1982, 17, 1499-1504.
Inorganic Chemistry, Vol. 46, No. 4, 2007 1403

Opalka et al.
Table 1. Best Structural Models Tested for Na2LiAlD6 Powder
Neutron Diffraction at 8 Ka
rotation
model
space
group
distance
Al-D (Å)
no. atomic
parameters
ø2
I a0b+c-
C2/c
P21/n
P42/n
P2/c
P1h
2.07
2.08
2.03
1.98
1.768
1.765
1.763
1.770
11
12
10
22
48
II a+b-b-
III a+a+c-
IV a+b+c-
V a+b-c-
a Models based on ordered perovskite structures from Woodward, 1997,
rotation mode according to Glazer, 1975; space group, ø2, average Al-D
distance and number of atomic parameters are given.
Figure 2. Na2LiAlH(D)6 powder neutron diffraction (PND) structural
analyses at 8 and 295 K.
Figure 4. Refinement of 8 K Na2LiAlH(D)6 structure.
space group P21/n, is shown in Figure 4. Except for well-
known instrumental peaks from the cooling system at 31 and
46°, the fit is acceptable. The last two models have a
significantly higher number of refined atomic parameters
without achieving the corresponding improvement in the fit
and can hence be rejected. Models I-III have the same fit
of the PND data, approximately the same number of atomic
parameters, and very similar average Al-D distances that
are close to the distances at room temperature. From PND
measurements it was confirmed that the transition takes place
above 110 K. Therefore, SR-PXD data at 110 K was used
in an attempt to distinguish between the models. But no
splittings of the cubic reflections, i.e., the reflections of the
Fm3hm model, were observed. From PND and SR-PXD, it
appeared that Na2LiAlD6 at low temperature assumed one
of ordered perovskite structures in space group C2/c, P21/n,
or P42/n described by Woodward,32 but it was not possible
to conclude which one best described the crystal structure.
The first three tilt structures that gave the best refinement
fits (cf. Table 1) were reviewed to prepare for Phonon direct
method lattice dynamic simulations. First, full ground state
minimizations were made of the newly identified C2/c and
P21/n monoclinic structures, and the P42/n tetragonal struc-
tures. The energies and the structural properties are compared
to the previous results calculated for the P21/c, Fm3hm, and
Immm structure types8 in Table 2. The P21/n structure is
related to the previously determined P21/c structure, where
each structure corresponds to a different unit cell representa-
tion of space group no. 14. It can be seen that with the
exception of the Fm3hm and Immm structures, the energies
Figure 3. Crystal structure of Na2LiAlD6 at room temperature. AlD6
octahedra (blue) and LiD6 octahedra (green) form a corner-sharing network
where each corner is shared between two octahedra. Na atoms (red) are in
interstitial positions surrounded by 12 D atoms.
sharing AlD6 and LiD6 octahedra of different sizes, where
each D corner is shared between AlD6 and LiD6 octahedra,
with 180° Al-D-Li bond angles, as shown in Figure 3.3
These octahedra may be rotated along the three axes and
may be tilted in either a positive or negative mode, where
adjacent octahedra layers are rotated in the same or opposite
direction, respectively. Glazer31 and later Woodward32 studied
all the different possible tilt systems, both for normal
perovskites (ABX3) and for ordered perovskites (A2BB′X6).
They have given the unit cell size, space group, and
approximate atomic positions of the different tilt systems.
There are 23 tilt systems with 15 different models. By
looking at the unit cell size and extinctions of the space
groups, seven of the models were eliminated. The remaining
eight models were checked by Rietveld refinements of the
PND data, and five turned out to have approximately the
same reliability factor ø2 of 1.98-2.08, as shown in Table
1. The Rietveld fit of one of the models, i.e., model II with
(31) Glazer, A. M. Acta Cryst. A 1975, 31, 756-762.
(32) Woodward, P. M.; Acta Cryst. B 1997, 53, 32-43.
1404 Inorganic Chemistry, Vol. 46, No. 4, 2007

The Na-Li-Al-H System
Table 2. Ground State Minimized Na2LiAlH6 Candidate Structures
lattice paramete rs
space
electronic energy
volume
group (kJ mol-1 atom-1) a (Å) b (Å) c (Å) â (deg) (Å3/atom)
Immm
Fm3hm
C2/c
P42/n
P21/n
P21/c
-289.2
-300.2
-300.5
-300.5
-300.6
-300.6
5.405 5.410 7.647
7.338 7.338 7.338
7.325 7.371 7.359
7.356 7.356 7.370
5.194 5.208 9.006 125.20
5.165 5.251 7.339 90.03
11.16a
9.88a
9.89
90.01
9.97
9.95
9.95a
a Reference 8.
are very close to one another and are not distinguishable
within the limits of accuracy for VASP ground state
calculations. These results can be interpreted to imply that
there may be a number of energetically equivalent crystal
structures accessible at the ground state.
Figure 5. Phonon dispersion plots for the Na2LiAlH6 (a) Fm3hm and (b)
P21/n structures.
Phonon direct method lattice dynamic predictions were
made on the highest symmetry Fm3hm, P42/n, and P21/n
minimized structures. This enabled the comparison of the
phonon dispersion relations of the original refined Fm3hm
structure with two of the most stable, energetically equivalent
simulated structures. The phonon dispersion relations for the
Fm3hm and P21/n phases are shown in Figure 5a-b. Both
the Fm3hm and the P42/n dispersion relations exhibited
significant imaginary frequencies representing instabilities
to transformation to other phase structures. The Fm3hm
structure dispersion had an unstable T1g acoustical mode at
the γ point. The P42/n structure had two imaginary Bg and
Ag Raman-active branches. The P21/n structure lacked
imaginary dispersion frequencies and was hypothesized to
be the most stable ground state phase.
Figure 6. The maximal group-subgroup relations connecting the cubic
Fm3hm structure with the monoclinic P21/n and C2/c structures through the
P4/mnc structure, which is the result of an instability of the Fm3hm structure
at the X point.
These observations brought forth to question the relation-
ship between the predicted stable P21/n ground state structure
and the experimentally observed Fm3hm room-temperature
structure. The first scenario postulated was that the Fm3hm
structure could undergo one or more second-order transfor-
mations with decreasing temperature to form the P21/n
structure. One possible pathway could be a multistep
transformation of Fm3hm to P42/n structure, followed by
transformation to the P21/n structure. However, animation
with the Medea Phonon module of the Fm3hm instability at
the Γ point led to a P1h structure. Likewise, animation of the
instabilities at the W, X, K, and L points led to a structures
in the space groups P2, P4/nmc, Pm, and R3, respectively.
Following an established approach for the investigation of
phase transformations,33 the Bilbao Crystallographic Server
Subgroupgraph program34 was used to determine whether
the P2, P4/mnc, Pm, and R3 spacegroups and other space-
groups could be intermediate subgroups between the Fm3hm
spacegroup and the P21/n, C2/c, or P42/n structures. Indeed,
such a direct transformation pathway was found between the
Fm3hm and the P21/n and C2/c structures; this is illustrated
in Figure 6. The figure shows that Fm3hm is connected to
both the monoclinic P21/n and C2/c structures via the X
instability, through the P4/mnc structure. This means that
Figure 7. Thermodynamic predictions determined from direct method
lattice vibrational calculations.
these monoclinic structures are closely related to the instabil-
ity of the cubic structure, which may explain why the latter
is not stable at low temperatures.
The predicted GHSER values for the P21/n Na2LiAlH6
phase are compared to those for the Fm3hm phase in Figure
7, where only the integrated phonon density of state values
over frequencies g0 were used to predict the Fm3hm phase
thermodynamic values. The P21/n phase was predicted to
be significantly more stable than the Fm3hm phase at all
temperatures. A crossover in the GHSER values, which
would indicate a transformation from the P21/n phase to the
Fm3hm phase with increasing temperature, was not predicted.
A second simpler scenario relating the phase structures is
even more credible. The P21/n and other simple tilt structures
(33) Perez-Mato, J. M.; Aroyo, M.; Garc´ıa, A.; Blaha, P.; Schwarz, K.,
Schweifer, J.; Parlinski, K. Phys. ReV. B 2004, 70, 214111/1-14.
(34) Ivantchev, S.; Kroumova, E.; Madariaga, G.; Perez-Mato, J. M.; Aroyo,
M. I. J. Appl. Cryst. 2000, 33, 1190-1191.
Inorganic Chemistry, Vol. 46, No. 4, 2007 1405

Opalka et al.
Table 3. Summary of Reported Na2LiAlH(D)6 Mechanochemical
Syntheses
anneal
anneal
reactants
T (K)
P (bar)
a
LiH + 2NaH + AlH3
LiAlD4 + 2NaHb
LiAlH4 + 2NaHc
373
423
-
200
50
-
d
c
LiH + NaH + NaAlH4
LiH + NaH + NaAlH4
-
-
-
453
453
-
30
80
b
LiAlD4 + 2NaAlD4
e
LiH + 2NaAlH4
a Reference 35. b Reference 3. c Reference 5. d Reference 2. e Reference
4.
could be actually comprised of two (or more) mirror-
symmetry equivalent structures, each where the AlH6 octa-
hedra are tilted off-center from the Fm3hm cubic structure in
opposite directions. With the availability of kinetic energy
with increasing temperature, the energy barriers to tilting
could be overcome, and the octahedra could rapidly tau-
tomerize between these equivalent structures. During the
crystallographic analyses at elevated temperatures, these
equivalent oscillations could be time-averaged to appear as
the intermediate cubic structure. In order to explore this
possible phenomenon, the Fm3hm structure was recast in
P21/n symmetry, and a tilting pathway with three intermedi-
ate steps was constructed between the two structures. By
recasting the Fm3hm structure in the P21/n symmetry, the unit
cell and atom positions were as similar as possible to the
P21/n structure. This gave an almost perfect cancellation of
any systematic errors in the calculations of structures along
the pathway, yielding accurate relative energies. The total
energy calculated for each step along the tilting pathway
showed a progressive increase with no intermediate barriers,
starting from the P21/n and ending with the Fm3hm structure.
To oscillate from one P21/n structure to the mirror symmetry
equivalent structure, the octahedra would have to pass
through a ground state intermediate cubic barrier of 0.4 kJ
mol-1 atom-1, the difference in energy between the P21/n
and Fm3hm structures. Any thermal energy or even the zero-
point vibrational energy would be sufficient to overcome this
barrier, as mentioned above, so for such a small barrier the
structure can be characterized as the average of the tautomers,
which is the Fm3hm structure. The structure at any reasonable
temperature may actually be the tautomer average of a
number of energetically similar structures that represent
different tilting of the AlH6 octahedra (such as the most
favorable structures listed in Table 2).
Figure 8. Predicted Na2LiAlH6 synthesis reactions.
predicted thermochemical equilibrium constants, Keq, for the
related Na2LiAlH6 synthetic reactions versus temperature
LiH + 2NaH + AlH3 ) Na2LiAlH6
LiAlH4 + 2NaH ) Na2LiAlH6
(1)
(2)
(3)
(4)
LiH + NaH + NaAlH4 ) Na2LiAlH6
1/3Li3AlH6 + 2/3Na3AlH6 ) Na2LiAlH6
LiAlH4 + 2NaAlH4 ) Na2LiAlH6 + 2Al + 3H2 (5)
LiH + 2NaAlH4 ) Na2LiAlH6 + Al + 3/2H2 (6)
are plotted in Figure 8. The higher the Keq, the greater the
reaction equilibrium shifts to the products. As expected, in
reactions 1 and 2 with the metastable AlH3 and LiAlH4
precursors were predicted to have a greater Na2LiAlH6 yield.
However, the Keq for these reactions decreases with increas-
ing temperature, and reaction 5 is predicted to predominate
at temperatures greater than ∼373 K (∼100 °C). These
predictions show why the reaction 5 using NaAlH4 and
LiAlH4 precursors employed by Brinks et al.3 required a
453 K anneal to increase the Na2LiAlH6 fraction. These
results demonstrate the potency of the predictive thermody-
namic methodology for designing the best experimental
conditions for synthesis of a target phase.
The thermochemical predictions provided insight into how
various possible dissociation mechanisms may be interrelated
with changing temperature. Hydride dissociation reactions
are commonly described in terms of reaction dissociation
enthalpies, ∆Hdis, determined from the slope of the Arrhenius
plot of the equilibrium H2 pressure formed upon dissociation
versus the reciprocal of the equilibrium temperature, accord-
ing to the van’t Hoff equation: ln PH2 ) ∆Hdis/RT + ∆S/R,
where P is the equilibrium H2 pressure at temperature T, S
in the dissociation entropy, and R is the universal gas
constant. The predicted ∆Hdis for Na2LiAlH6 and Na3AlH6
dissociation reactions per dissociation of 1 mole H2
Thermodynamic descriptions generated from the Phonon
direct method were used to evaluate compound formation
and dissociation reactions within the Na-Li-Al-H system.
The various mechanochemical routes reported in the literature
for the synthesis of the Na2LiAlH6 phase2-5,35 are sum-
marized in Table 3. Many of these syntheses used annealing
treatments to increase the detectable yield of Na2LiAlH6. The
(35) Anton, D. L.; Opalka, S. M. VI.A.4 Complex Hydride Compounds
with Enhanced Hydrogen Storage Capacity; DOE Hydrogen Program
2005 Annual Progress Report; U.S. Department of Energy: Wash-
ington, DC, 2005 (http://www.hydrogen.energy.gov/pdfs/progress05/
vi_a_4_anton.pdf).
1406 Inorganic Chemistry, Vol. 46, No. 4, 2007

The Na-Li-Al-H System
Table 4. Comparison of Predicted and Measured Dissociation Energies
dissociation enthalpy ∆Hdis (kJ/mol)
2/3Na2LiAlH6 ) 4/3NaH + 2/3LiH + 2/3Al + H2 (7)
2/3Na3AlH6 ) 2NaH + 2/3Al + H2
(8)
predicted
measured
2/3Na2LiAlH6 )> 4/3NaH + 2/3LiH + 2/3Al + H2
are compared to ∆Hdis values determined from pressure
composition isotherm measurements4,5,36 in Table 4. The
predicted ∆Hdis values were smaller, but were within 10 kJ/
mol of the reported measured equilibrium values. A residual
error of up to10 kJ/mol in the predicted ∆Hdis can impact
the predicted dissociation pressure by a factor of up to 25 at
373 K.
47.5
56.4a
53.5b
2/3Na3AlH6 )> 2NaH + 2/3Al + H2
44.8
47c
a Reference 4. b Reference 5. c Ref 36.
reacted. The same effect was seen, but too less degree, with
pure Na2LiAlH6: Traces of Na3AlH6 were seen in the
diffraction diagrams at intermediate temperatures below the
main dissociation of Na2LiAlH6. This proved experimentally
that at lower temperature, Na3AlH6 was more stable than
Na2LiAlH6, whereas at higher temperature Na3AlH6 was less
stable than Na2LiAlH6. In addition, it showed that the metal
ions had sufficient mobility for such a reaction to take place
at relatively low temperature.
Phase diagram predictions were made to assess the relative
stability of the different phases in the quaternary Na-Li-
Al-H system with composition, temperature, and pressure.
An excellent perspective was obtained from Na-Li-Al
ternary sections calculated at constant H2 activity, temper-
ature, and pressure. Ternary sections generated with the fixed
conditions (a) 100 bar, 300 K (27 °C), and 3.9 × 10-3 H2
activity and (b) 100 bar, 400 K (127 °C), and 1.0 × 10-3 H2
activity are shown in Figure 10a,b. The lines in these
diagrams demarcate the regions where common sets of
phases coexist. The Na2LiAlH6 phase is shown to coexist
with the ternary alanate and the elemental hydride phases is
predicted over a wide compositional range in these diagrams.
The phase equilibria were systematically surveyed over
varying conditions to identify the composition, temperature,
and pressure conditions for the optimum reversible hydrogen
retrieval in the Na-Li-Al-H system. The optimum com-
position contained the elemental stoichiometry: {2 Na:1 Li:2
The differences between the predicted and measured values
were most likely due to the inherent inaccuracies in the
ground state minimizations with the density functional
exchange-correlation GGA approximation, and/or the har-
monic approximation which does not account for the possible
anharmonic free rotational modes of the AlH6 octahedra.
Such rotations would be described by 3 of the 27 total
degrees of freedom in the Na2LiAlH6 structures, and thus
affect a relatively small proportion of the vibrational
thermodynamic contributions. An upper bound to the error
from using the harmonic approximation for the rotations of
the AlH6 groups is given by assuming that they are free rotors
and comparing the thermodynamic functions to those cal-
culated for the approximately 3 THz harmonic oscillators
from the harmonic approximation. The error in the enthalpy
and free energy is no more than 6 kJ/mol over the
temperature range considered. This is an upper bound and
the actual error due to the harmonic approximation is
probably considerably less.
The van’t Hoff plot for the most favorable predicted
dissocation reactions are shown in Figure 9, in terms of
equivalent starting amount of Na2LiAlH6 (not per equivalent
mole of H2 generated), for ease of comparison. The expected
dehydrogenation reaction to form binary hydrides, aluminum
and 3.5 wt. % H2 [reaction 7], is predicted to be favorable
at temperatures above 360 K (87 °C). The predicted
dissociation curve traverses the lower left-hand corner of the
window of reversibility, generating a lower dissociation
pressure with respect to the dissociation reaction for the
equivalent amount of Na3AlH6 (the comparison can be made
by referring to Table 4). At temperatures below 360 K, a
higher dissociation H2 pressure and hence greater favorability
is predicted for the reaction
2/3Na2LiAlH6 ) 4/9Na3AlH6 + 2/3LiH + 2/9Al + 1/3H2
(9)
which releases only 1.2 wt. % H2.
Experimental observations corroborated the predicted
relationship between the competing Na2LiAlH6 dissociation
reactions 7 and 9. In situ PXD showed that for Ti-enhanced
Na2LiAlH6 heated in vacuum at 1 K/min, Na2LiAlH6 was
partially dissociated via Na3AlH6 before Na2LiAlH6 was
thermally dissociated to NaH, LiH and Al. The Na3AlH6
formed was dissociated before the remaining Na2LiAlH6
Figure 9. Predicted van’t Hoff plot for most favorable Na2LiAlH6
dissociation reactions, compared for equivalent starting amount of Na2-
LiAlH6. The in situ window of reversibility is defined by the practical 1-100
bar pressure range and 25-100 °C temperature range for hydrogen storage
system operation.
(36) Bogdanovic, B.; Brand, R. A.; Marjanovic, A.; Schwickardi, M.; Tolle,
J. J. Alloys Comp. 2000, 302, 36-58.
Inorganic Chemistry, Vol. 46, No. 4, 2007 1407

Opalka et al.
Figure 10. Na-Li-Al ternary sections from thermodynamic property predictions (a) at 100 bar, 300 K and 3.9 × 10-3 H2 activity with phase mixture
regions (i) Al, LiH, Na2LiAlH6, (ii) NaH, LiH, Na2LiAlH6, (iii) NaH, Na3AlH6, Na2LiAlH6, and (iv) Al, Na3AlH6, Na2LiAlH6 and (b) at 100 bar, 400 K
(127 °C) and 1.0 × 10-3 H2 activity with phase mixture regions (v) Al, NaAlH4, Li3AlH6, (vi) LiH, NaAlH4, Li3AlH6, (vii) LiH, NaAlH4, Na2LiAlH6, (viii)
LiH, NaH, Na2LiAlH6, (ix) NaH, Na3AlH6, Na2LiAlH6, and (x) NaAlH4, Na3AlH6, Na2LiAlH6.
Al:9 H}. Under the conditions of 300 K and 100 bar, this
composition is predicted to form fully hydrided NaAlH4 and
LiH phases, in a ratio of 2:1. The predictions show that LiH
destabilizes the dissociation of NaAlH4 to form the more
favorable Na2LiAlH6 phase [reaction 6], increasing the H2
plateau pressure approximately 1 order of magnitude to 11
bar at 306 K (33 °C) compared to the first dissociation
36
reaction of pure NaAlH4
NaAlH4 ) 1/3Na3AlH6 + 2/3Al + H2
(10)
However, the benefits for destabilization are compromised
by the extra weight of LiH which dilutes the total capacity
of NaAlH4, and the smaller amount of H2 released in the
first dissociation step: 2.6 wt. % H2 for reaction 6 versus
3.7 wt. % H2 for reaction 10. At temperatures higher than
360 K at 1 bar pressure, the dissociation of the Na2LiAlH6
phase [reaction 7] has a greater equilibrium constant than
the dissociation of the equivalent amount of pure Na3AlH6
[reaction 8]. The amount of hydrogen released is greater in
the Na2LiAlH6 decomposition reaction, 2.6 wt. % H2,
compared to that released in the second Na3AlH6 dissociation
reaction, 1.9 wt. % H2. It is envisioned that the two step
dissociation process could be used to extract 5.2 wt. % H2
reversibly from the optimum {2 Na:1 Li:2 Al:9 H} composi-
tion: first, degassing through reaction 6 by changing the
pressure from 100 to 1 bar at room temperature, and second,
degassing through reaction 7 by increasing the temperature
to g400 K at 1 bar pressure. This reaction pathway is
summarized schematically in Figure 11. In practice, often
even higher temperatures and/or lower pressures are required
to overcome kinetic barriers and to achieve acceptable
hydrogen discharge rates. The feasibility of the forward
reaction scheme has been experimentally confirmed in recent
synthetic studies;35 however, the reversibility and kinetics
of these coupled reactions await further experimental ex-
amination.
Figure 11. Reaction schematic for maximum theoretical H reversibility
in Na-Li-Al-H system.
behavior of the Na2LiAlH6 phase. A low-temperature P21/n
monoclinic structure was jointly identified through crystal-
lographic refinement and direct method lattice dynamics to
be a suitable stable ground state structure for thermodynamic
property predictions. Two feasible transformation pathways
were proposed to relate this structure to the less stable room
temperature observed Fm3hm structure. First, the Fm3hm
structure may directly transform to the P21/n structure along
the X instability, passing through an intermediate P4/mnc
structure. Second, the Fm3hm structure could actually be the
time-average of a structure that rapidly oscillates between
one or more pairs of mirror-symmetry perovskite distortions,
formed by tilting of the AlH6 octahedra. The thermodynamic
predictions enabled the thermodynamic modeling of the Na2-
LiAlH6 syntheses and dehydrogenation reactions, generating
unique insights into competing reaction mechanisms. The
calculated bialkali dissociation pressure relationships were
compared with those determined from pressure-composition-
isotherm experiments to validate the predicted thermody-
namic properties. The predictions of varying Na2LiAlH6
dehydrogenation behavior with temperature were substanti-
ated by experimental observations and were used to aid in
their interpretation. Gibbs energy minimization and phase
diagram predictions were then used to determine the reaction
Conclusion
Combined experimental and theoretical methodologies
were used to investigate the structure and transformation
1408 Inorganic Chemistry, Vol. 46, No. 4, 2007

The Na-Li-Al-H System
pathways for the maximum theoretical capacity of 5.2 wt.
% H2 that could be achieved with the Na-Li-Al-H system.
Research Council through the Nanomat program. The
original integrated experimental-theoretical Na-Li-Al-H
thermodynamics were generated in collaboration with C. Qiu
and G. B. Olson of QuesTek Innovations, LLC. The skillful
assistance from the project team at the Swiss-Norwegian
beam line, ESRF, is also gratefully acknowledged.
Acknowledgment. S. M. Opalka acknowledges support
by the U.S. Department of Energy Contract DE-FC04-
02AL67610, and team members: X. Tang, D. A. Mosher,
and B. L. Laube and former project leader D. L. Anton. O.
M. Løvvik acknowledges support from the Norwegian
IC062032E
Inorganic Chemistry, Vol. 46, No. 4, 2007 1409