Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations Ole Martin Løvvika) University of Oslo, Centre for Materials Science and Nanotechnology, 0318 Oslo, Norway Ole Swang SINTEF Materials and Chemistry, N-0314 Oslo, Norway Susanne M. Opalka United Technologies Research Center, East Hartford, Connecticut 06018 (Received 14 April 2005; accepted 10 June 2005) The alanates (complex aluminohydrides) have relatively high gravimetric hydrogen density and are among the most promising solid-state hydrogen-storage materials. In this work, the crystal structure and electronic structure of pure and mixed-alkali alanates were calculated by ground-state density-functional band-structure calculations. The results are in excellent correspondence with available experimental data. The properties of the pure alanates were compared, and the relatively high stability of the Li3AlH6 phase was pointed out as an important difference that may explain the difficulty of hydrogenating lithium alanate. The alkali alanates are nonmetallic with calculated band gaps around 5 eV and 2.5–3 eV for the tetra- and hexahydrides. The bonding was identified as ionic between the alkali cations and the aluminohydride complexes, while it is polar covalent within the complex. A broad range of hypothetical mixed-alkali alanate compounds was simulated, and four were found to be stable compared to the pure alanates and each other: LiNa2AlH6, K2LiAlH6, K2NaAlH6, and K2.5Na0.5AlH6. No mixed-alkali tetrahydrides were found to be stable, and this was explained by the local coordination within the different compounds. The only alkali alanate that seemed to be close to fulfilling the international hydrogen density targets was NaAlH4. I. INTRODUCTION The search for solid-state hydrogen-storage materials is a crucial part of the path toward implementation of a hydrogen economy.1–7 Alkali alanates [complex hydrides with the formula MnAlH(n+3), where M is an alkali element: Li, Na, K, . . .] are among the most promising materials for this application, with high hydrogen density available at moderate conditions.8–13 The fully hydrogenated alkali alanate phase, the monoalkali aluminium tetrahydride, MAlH4 desorbs hydrogen stepwise through a series of reactions 3MAlH4 → M3AlH6 + 2 Al + 3 H2 , M3AlH6 → 3MH + Al + 3/2 H2 3MH → 3M + 3/2 H2 , , (1) (2) (3) where M is Li, Na, or K. In the first reaction, the MAlH4 a) Address all correspondence to this author. e-mail: o.m.lovvik@fys.uio.no DOI: 10.1557/JMR.2005.0397 J. Mater. Res., Vol. 20, No. 12, Dec 2005 phase disproportionates to form the trialkali aluminium hexahydride, M3AlH6 phase, which during the second reaction disproportionates into the monoalkali hydride, MH. The last MH decomposition reaction takes place at temperatures too high to allow it to be counted as accessible for in situ mobile applications. The total and accessible hydrogen content of the alkali alanates are listed in Table I. A. Lithium alanate Since LiAlH4 is the lightest alanate, it is the most attractive candidate from a gravimetric point of view. The reversible hydrogenation of lithium alanate has been pursued by several groups,14–35 and it is now clear that, for lithium alanate, the decomposition of LiAlH4 [Eq. (1)] is spontaneous at room temperature, and the reaction is apparently exothermic.14,16,35 The decomposition reactions (1) and (2) are promoted by ball milling14,17,36 and catalysis.19,27 It has been claimed that the second step of the lithium alanate decomposition reaction [Eq. (2)] is reversible,19 but this has not been reproduced. It has rather been shown that, when doped with Ti, the © 2005 Materials Research Society 3199 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations TABLE I. Calculated properties of the monoalkali alanates. The calculated space group and hydrogen volumetric density are compared to experimental results. The total and accessible hydrogen weight percent are the amounts of hydrogen that theoretically may be extracted from the compound when going to the elements and to the alkali metal hydrides, respectively. The volume density is based on the accessible hydrogen content. Compound Space group H vol density (kg/m3) H wt% Formation enthalpy Reaction enthalpy Al–H distance (pm) Alkali–H distance (pm) H–Al–H angle (deg) Al–H coordination Alkali–H coordination LiAlH4 NaAlH4 KAlH4 Li3AlH6 This work Experimental This work Experimental Total Accessible P21/c P21/c 73.4 73.8 10.6 8.0 I41/a I41/a 72.0 72.6 7.5 5.6 Pnma Pnma 52.6 53.2 5.8 4.3 (kJ/mol H2) −55.5 −54.9 (kJ/mol H2) Minimum Maximum Minimum Maximum Minimum Maximum 669.7 162 165 185 198 107.8 111.5 4 5 670.2 164 164 241 241 107.5 113.5 4 8 equilibrium pressure for the first reaction [Eq. (1)] is higher than 99 bar at 58 °C, and for the second reaction [Eq. (2)], higher than 85 bar at 80 °C.35 Calculated crystal structures correspond well with experimental results.23,24,28,30,33 The calculated electronic density of states (DOS) show that LiAlH4 and Li3AlH6 are insulators with band gaps of around 5 and 3.5 eV.23,24,26,32 B. Sodium alanate For the sodium alanate phases, the forward and reverse reactions (1) and (2) are kinetically hindered under equilibrium temperature and pressure reaction conditions. It was discovered by Bogdanovic et al. that the reverse reactions (1) and (2) above are made possible at moderate equilibrium temperature and pressure conditions for sodium alanate by adding small amounts of a transition metal as a catalyst or dopant.8 Significant efforts have been invested to understand the role and nature of these additives,8–13,37–76 and it is now clear that, for sodium alanate Ti is the most efficient additive8,47 and exhibits the best performance when added as nanoparticles.13,53,56 The effect of Ti may be enhanced by simultaneously adding Zr37,41, Fe,72 or graphite.71 The kinetics may be further improved by mechanochemical treatment (e.g., ball-milling).40 Ti is zero-valent after ball-milling.60,62,68 A major part of the Ti may after a few hydrogenation cycles be found associated with Al,43,62,77 either as a solid solution in Al62,63 or in the form of amorphous TiAl3,54,55,68 dependent on the temperature used during the cycling process.74 Some studies show that the structural parameters of NaAlH4 change after adding Ti,45,73 others do not.61,63,75 Substitution of Ti in the NaAlH4 3200 Na3AlH6 K3AlH6 R3̄ R3̄ 57.2 56.8 11.2 5.6 P21/n P21/n 43.8 43.7 5.9 3.0 P21/n ... 31.8 ... 4.0 2.0 −70.0 −102.8 −69.9 −78.5 668.0 163 164 270 354 107.7 112.1 4 12 1274.0 174 175 190 209 87.1 93.0 6 6 1193.3 176 177 222 277 89.0 91.0 6 6–8 1168.4 178 179 258 363 89.8 90.2 6 6–11 sublattice is metastable, and the most stable substitution is for Al, near the surface.70 Some of the early discrepancies between various studies have been resolved as being due to differences in sample preparation, different stages in the hydrogenationdehydrogenation cycles, etc. There still are some contradictions in the literature, particularly as to whether or not the structural parameters change after Ti addition. This is an important issue, since it has been shown that the structural parameters would change significantly upon substitutional doping and would be analytically detectable even down to levels as low as 3% Ti doping.70 In any case, there is still a need for better understanding of the doping/catalysis process, since this would guide a systematic theoretical search for materials and additives with effective properties for reversible hydrogen storage. A number of theoretical studies have been devoted to the sodium alanates.24,26,28,31,32,61,69,70,78–84 The calculated crystal structures24,26,28,32,61,69,78,83 and formation enthalpies61,78,81 at 0 K are in good correspondence with experimental results. The uncorrected electronic density of states shows that NaAlH4 and Na3AlH6 are both nonmetallic with band gaps of around 5 and 3 eV at the generalized gradient approximation (GGA) level,24,28,32,61,82,83 increasing to 6.9 and 4.6 eV when adding quasiparticle GW corrections85,86 to the band structure.82 The bonding is ionic between Na+ and the AlH4− tetrahedra in NaAlH4. The bonding within the tetrahedron has been suggested to be both ionic79 and covalent,24 but it may probably best be described as covalent with a strong ionicity (polar covalent).61,82 Phonon calculations on NaAlH4 have shown that the AlH4− anion remains stable up to the melting point,75 and J. Mater. Res., Vol. 20, No. 12, Dec 2005 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations thermodynamic functions are in good agreement with experimental ones.83 The trisodium aluminium hexahydride phase has been shown by experiment and calculations to exist as two displacively related, allotropic phases, referred to as ␣ and . The low-temperature ␣ disordered phase has a monoclinic P21/n symmetry. The high-temperature  phase has an orthorhombic Immm symmetry, analogous to the well-known cryolite, Na3AlF6, perovskite structure.78,87 C. Potassium alanate Potassium alanate is significantly heavier than lithium or sodium alanate, but compared to interstitial transition metal hydrides, it may still be of interest as a mediumtemperature hydride. It is much less studied than the lighter alkali alanates,31,32,88–90 but it is known that hydrogenation of potassium alanate is reversible even without a catalyst, and the reactions take place between 250 and 340 °C. 88 The calculated crystal structure of KAlH426,90 is in good correspondence with the experimental one.89 The crystal structure of K3AlH6 is not known, but it has been predicted to be in the P21/n space group.31,32 The electronic DOS shows that KAlH4 and K3AlH6 are both insulators, with band gaps of 5.590 and 3 eV.32 D. Mixed-alkali alanates The mixed alanate LiNa2AlH6 was synthesized for the first time in 1982,91 and it was demonstrated to be a reversible hydrogen storage material already in the seminal paper of Bogdanovic et al.8 It was later claimed that mixtures with the nominal compositions Li1.3Na1.7AlH6 and Li1.5Na1.5AlH6 also could absorb hydrogen reversibly,92 but this has not been followed up by diffraction measurements—the compositions could thus possibly be mixtures of LiNa2AlH6 and Li3AlH6, where only the former compound takes part in the hydrogenation. A structure with the stoichiometry Li1.5Na1.5AlH6 proved stable compared to the unmixed hexahydrides in a computational study, 81 but it is unstable compared to LiNa2AlH6 and Li3AlH6. So far no mixed alanates in the tetrahydride system [in the form MxM⬘(1−x)AlH4] have been reported. The only published theoretical study on this, in which the authors investigated the possibility of a stable material with the composition NaxLi(1−x)AlH4 having the NaAlH4 structure, failed to identify any stable compounds.81 The crystal structure of LiNa2AlH6 was suggested to be cubic face-centered-cubic (fcc),91,93 and it was completely resolved in the Fm3̄m space group by a combined synchrotron x-ray and neutron powder diffraction study.35 Calculations show, however, that a monoclinic structure is more stable at 0 K.31,81 This is an effect of temperature, and phonon calculations indicate that the cubic structure may be more favorable at higher temperatures.94 The calculations are supported by a recent diffraction study that shows distortions from cubic symmetry in LiNa2AlH6 at low temperature.95 The addition of Ti has the same positive effects on the hydrogenation kinetics of LiNa2AlH6 as on that of NaAlH4.8 It has later been shown that the addition of La2O3 gives similar results.96,97 When potassium was included in the search for mixed alanates, three other compounds were found to be stable compared to the pure alanates: K2LiAlH6, K2NaAlH6, and KNa2AlH6.31,32 The latter was not stable compared to K2NaAlH6 and Na3AlH6, however, and a recent experimental study managed to synthesize K2LiAlH6 and K2NaAlH6, but not KNa2AlH6.98 K2NaAlH6 was synthesized for the first time in 1987.99 E. This paper Our objective for this paper was to summarize our modeling investigations of the bulk alanates in the ground state,29–32,70 with an emphasis on results that were previously unpublished. Here, we first report on the simulation of already known alkali alanate crystal structures to validate this relatively new methodology for identifying correct structures, and also, to set the stage for the simulation of unknown mixed alkali compounds. These calculations are the required first step to prepare for the prediction of unmeasured, or difficult to measure, alkali alanate finite temperature thermodynamic properties by first-principles methods. We have extended our previous search for mixedalkali alanates,31,32 and in the following, we present a broad range of hypothetical hexahydride mixtures [of the form MxM⬘(3−x)AlH6]. Additionally we have performed a comprehensive search for mixed-alkali tetrahydrides [of the form MxM⬘(1−x)AlH4] based on Li, Na, and K. The paper starts with a presentation of the theoretical tools used in the calculations, followed by results for the pure and unmixed alkali alanates. A summary and assessment of the future of alanates as hydrogen-storage materials conclude the paper. II. METHODOLOGY Most of the calculations in this work have been performed using the Vienna ab initio simulation package (VASP),100,101 employing the projector augmented wave (PAW) method102 at the GGA level.103 The spacing of the integration grid in k-space was at the most 0.05 Å−1, in many of the cases below 0.02 Å−1. Self-consistency of the electronic density was defined to be reached when the total energy of two consecutive runs differed by less than 0.01 meV. The single-particle orbitals were smeared by a J. Mater. Res., Vol. 20, No. 12, Dec 2005 3201 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations Gaussian convolution with a width of 0.2 eV in most of the cases. The cutoff energy for the plane wave expansion was set to 780 eV. The overall convergence of the total energy with respect to the mentioned sources of error was in all cases on the order of 1 meV. Ionic relaxations were performed by the residual minimization method direct inversion in iterative subspace implementation of the quasi-Newton algorithm. In some difficult cases a conjugate-gradient algorithm was used in combination with the quasi-Newton method. The criterion for relaxation was that the forces be less than 0.05 eV/Å. The search for unknown crystal structures was performed by using a number of representative crystal structures from equivalent compounds as input structures for the ionic relaxation. The input structures were chosen to represent as large a variation in coordination numbers and crystal systems as possible. The full minimization method used allowed simultaneous relaxation of ionic coordinates, unit cell size, and shape. Thus, both known and unknown crystal structures were surveyed, including a large part of the realistic crystal structure space in the search. The input structures for the full structural minimization calculations were based on NaClO 4 (F4̄ 3m), NaAlH 4 (P4/mmm), KAuCl 4 (Pc), RbAlF 4 (Pmmn), LiBH 4 (P6 3 mc), KAlD 4 (Pnma), NaAlD 4 (I4 1 /a), LiAlD 4 (P2 1 /c), KAlF 4 (P2 1 /m), NaAlCl4(P212121), KAlCl4(P21), and NaGdCl4(P1̄) for the tetrahydrides. The hexahydrides were relaxed from the following input models: K3MoF6(Fm3m), Ti3NiS6(R3̄), Na 3 AlF 6 (Pmmn), Li 3 AlF 6 (Pna2 1 ), Na 3 AlD 6 (P2 1 /n), K 3 Fe(CN) 6 (P2 1 /c), and triclinically distorted Ti3NiS6(P1). References to the model structures have been published elsewhere.30 The crystal orbital overlap populations and Hirshfeld charges HihHH were calculated using ADF-BAND.104,105 This program uses linear combinations of atomic orbitals as basis functions. We used combinations of Slater-type and numerical atomic orbitals as basis sets, one of each for valence electrons and in addition two Slater-type orbitals as polarization functions. The electronic structure was calculated self-consistently within the local density approximation, and the zeroth order regular approximation was used to include relativistic effects. The number of k-points and integration accuracy were chosen to give total energy convergence within 10 meV per unit cell. III. RESULTS A. Monoalkali alanates The crystal structures of almost all the lightest MAlH4 and M3AlH6 phases, based on Li, Na, and K, are by now well known, except that of K3AlH6. The crystal structures have been experimentally described in detail elsewhere.20,21,89,106,107 We have listed our calculated structures compared to 3202 the experimental ones in Table I. In all cases, the agreement with experiment is very good: The correct space group is reproduced, and the calculated volume is within some 1% of the experimental volume. The calculated atomic positions (not shown here) are also very close to the experimental ones. We should not expect any better agreement than this, since our calculations are performed at 0 K, while the experiments are at finite temperatures, in many cases at room temperature. We also did not include zero-point motion in any of our calculations, which has been shown to change the lattice constants of NaAlH4 by up to 2%.83 It is interesting to see that the Al–H distance and coordination are almost constant both in the tetrahydrides (four-coordinated with Al–H distances of around 163 pm) and the hexahydrides (six-coordinated with Al–H distances of around 177 pm). This shows that the AlH4 tetrahedra and AlH6 octahedra are quite stable units independent of the local environment, consistent with the stability of the AlH4 tetrahedra observed at elevated temperatures in NaAlH4 by Raman scattering.75 This is also supported by the calculated H–Al–H angles within the aluminohydride complexes, which consistently deviate from perfect polyhedra. The only apparent trend is that the deviation from perfect octahedra in the hexahydrides decreases when the size of the alkali cation increases; thus the angles within the AlH6 octahedron in K3AlH6 differ only by up to 0.2° from 90°, while they differ by up to 3° in Li3AlH6. Apart from this, all the differences in local environment between the different alkali alanates are manifested by the alkali atom and the aluminohydride complex functioning as independent units. The alkali atom-H coordination naturally increases when the radius of the alkali atom increases, and so does the alkali-H distance. If the bonding is assumed to be ionic, the Shannon ionic radii may be good measures on the atomic sizes within the structure.108,109 When the tetrahedral and octahedral Shannon ionic radii of Al are subtracted from the Al–H distances, we obtain an effective ionic radius of hydrogen of 109 to 111 pm when tetrahedrally coordinated to Al (in the tetrahydrides) and between 107 and 111 pm when octahedrally coordinated (in the hexahydrides). That is, the hydrogen size is virtually unchanged when going from tetrahedral to octahedral coordination in this picture. If these radii are used for the hydrogen atom, we similarly obtain effective radii for the alkali atoms from the alkali–H distances. In LiAlH4 the radii obtained in this way vary from 76 to 87 pm, that is between the Shannon ionic radii of tetrahedrally coordinated Li (73 pm) and octahedrally coordinated Li (90 pm), consistent with the 5-fold coordinated Li in this structure. For NaAlH4, however, the effective ionic radius of Na is between 130 and 132 pm, significantly higher than the tetrahedral (113 pm) and octahedral (116 pm) Shannon ionic radii. The difference J. Mater. Res., Vol. 20, No. 12, Dec 2005 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations is even larger in KAlH4, with estimated effective ionic radii between 161 and 243 pm, compared to the octahedral Shannon ionic radius of 152 pm. Since the latter structure has 11-coordinated K, we could expect slightly larger radii than the octahedral Shannon ionic radius. It is nevertheless clear that the Li ion takes relatively less space in the structure than what should be expected from the ionic radius and the structures of the other alanates. This may partially be explained by the relatively weak ionicity and large Li–H overlap in Li alanate, as we shall see later. Despite this, the volume of LiAlH4 is only about 2% less than that of NaAlH4. On the other hand, the volume of Li3AlH6 is 30% less than that of Na3AlH6. This reflects a packing of the ions in Li3AlH6 much more efficient than that in LiAlH4, and we shall see that this results in significant differences in the relative stability between the two phases of Li alanate compared to the heavier alanates. Morioka et al. observed K3AlH6 during the desorption of KAlH4,88 but this phase has not yet been crystallographically resolved. Earlier, we reported cell parameters and ionic positions of K3AlH6,32 but the structure has not been discussed in detail elsewhere. The predicted crystal structure of K3AlH6 shown in Fig. 1 is very similar to that of Na3AlH6, with the unit cell size being the largest difference. It is also slightly closer to the cubic Fm3̄m structure than Na3AlH6 in that the unit cell exhibits smaller deviations from cubic symmetry; the largest deviations of the lattice constants are 2.3% and 1.3%, and the largest deviations from 90° are 0.8° and 0.4° in Na and K alanate, respectively. Also, the AlH6 octahedra are less tilted (by around 6°) from the ideal perovskite structure. The potassium atoms in Wyckoff position d (two in the unit cell) are six-coordinated in quite regular octahedra (K–H distances between 258 and 260 pm and H–K–H angles between 87.4° and 92.6°). Those in Wyckoff position e (four in the unit cell) occupy interstitial voids between the AlH6 and KH6 octahedra and do not form regular polyhedra. They have coordination numbers of 8 or 11, depending on whether when hydrogen neighbours up to 310 or 360 pm distance are included, respectively. The calculated formation enthalpies and reaction enthalpies of the monoalkali alanates are also shown in Table I. The formation enthalpy Hform of the compound MnAlH(n+3) is defined as Hform[MnAlH(n+3)] ⳱ E[MnAlH(n+3)] – n E(M) – E(Al) – (n + 3)/2 E(H2) , (4) where E is the calculated total electronic energy of the alanate and the elements in their standard state, as calculated by VASP at 0 K. The reaction enthalpies are defined in accordance to the disproportionation reactions (1) and (2) for the tetrahydrides and the hexahydrides, respectively: Hreact(MAlH4) ⳱ E(MAlH4) – 1/3 E(M3AlH6) (5) – 2/3 E(Al) – E(H2) , Hreact(M3AlH6) ⳱ 2/3 E(M3AlH6) – 2 E(MH) – 2/3 E(Al) – E(H2) . FIG. 1. Crystal structure of K3AlH6. The octahedra are AlH6 complexes, and the balls are K atoms. (6) A more negative formation enthalpy is indicative of a more stable phase, relative to other candidate structures. Likewise, a more negative reaction enthalpy represents an increased favorability for disproportionation and release of H2. Table I shows that all the alanates are stable compared to the elements (negative formation enthalpy.) There is no clear trend in the formation enthalpy when going down the alkali series, but there is one important difference: Li3AlH6 has a much more negative formation enthalpy than the heavier hexahydrides. This may be an important reason why the reverse reaction (1) is difficult to achieve in the Li alanates: the hexahydride is too stable, making this reaction thermodynamically unfavorable. One reason why the Li hexahydride is so stable is simply the small size of the Li ion. This makes the hexagonal R3̄ structure accessible, which, as mentioned before, has significantly denser packing of the structure than the other hexahydrides, more than should be expected simply from the size of the ion. There is no such structure available for the Li tetrahydride, so the relative J. Mater. Res., Vol. 20, No. 12, Dec 2005 3203 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations stability of the two Li alanates is more in favor of the hexahydride than that of the heavier alanates. This picture of the difficult hydrogenation of LiAlH4 could of course be accompanied by kinetic barriers impeding the reaction. In addition, we should expect significant effects on the hydrogenation from finite temperature and zeropoint motion, but that is beyond the scope of this work. The total and local DOS of the tetrahydrides are shown in Fig. 2. (The corresponding plots for the hexahydrides are shown in Ref. 32). Many properties are similar between the three compounds, with some exceptions. The calculated band gap increases from Li alanate (4.7 eV) via Na (4.8 eV) to K (5.2 eV) alanate. Quasiparticle GW calculations have shown that the band gap of NaAlH4 is underestimated by 2–3 eV,82 so we may expect that all three compounds have quite large band gaps, at least on the order of 7 eV. The calculated band gaps of the hexahydrides are 3.1, 2.4, and 2.9 eV for the Li, Na, and K alanates. Most of the orbitals have distinctly different regions of available states in the valence and conduction bands, which means that the bonding is either ionic or polar covalent. The largest differences in the DOS are found between the two lightest alanates and K alanate. The band widths of K alanate are much smaller than those of the others, similar to what is seen in the hexahydrides.32 This means that we should expect smaller overlaps between the ions in K alanate than in the lighter alanates, and thus relatively lower density than expected from the size and mass of the atoms alone. There are no signs from the DOS that could explain the different hydrogenation behaviors of the alanates. Another quantitative measure of the bonding is the charge distribution and amount of overlap between the atoms. The calculated Hirshfeld charges and overlap populations are shown in Table II. Some clear trends may be read out of these data: while the charge on the alkali atom increases while moving from Li to K alanate, the charge on Al decreases. The charge on H also decreases slightly, so the charge on AlH4 balances that of the alkali atom. This is similar to what was found for the hexahydrides,32 and the same arguments apply here. Since the Li atom is small and has a very small degree of polarizability and low Pauling electronegativity, it is less cationic than the heavier atoms. The overlap population between Li and H in LiAlH4 is on the other hand higher than that in the heavier tetrahydrides, which shows a slightly stronger covalence in this compound. There is a significant overlap between Al and H in all the alanates, and this shows that the bonding between Al and H is covalent with strong ionicity, as it was put forward by Peles et al. for NaAlH4.82 It is interesting to see which orbitals contribute to the overlap, and the calculated crystal orbital overlap population (COOP) is presented in Fig. 4. The overlap is small or negative (antibonding) for the alkali–H interaction, while the Al–H overlap is large and bonding for all three alanates. The latter gives further strength to the Al–H interactions described as polar covalent. The s, p, and d orbitals of Al all contribute to the overlap: the s orbitals in the lower region of the valence band and the p and d orbitals in the upper. The electron localization functions (ELF)110 of the FIG. 2. Total and local DOS of the tetrahydrides LiAlH4, NaAlH4, and KAlH4. The DOS projected on alkali atoms, Al, and H are shown from left to right, with the total DOS together with H. s-, p-, and d-projected orbitals are shown as shaded gray, solid, and dotted curves, while the total DOS is plotted as a solid curve. The Fermi level is marked with a dashed line. 3204 J. Mater. Res., Vol. 20, No. 12, Dec 2005 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations FIG. 3. COOP between H and the metal atoms in the pure monoalkali aluminium tetrahydrides, calculated by ADF-BAND. The shaded gray, solid, and dotted curves represent metal s, p, and d orbitals. Positive and negative overlap represents bonding and anti-bonding, respectively. The Fermi level is marked with a dashed line. three monoalkali aluminium tetrahydrides are shown in Fig. 3. There is no localization between the atoms in any of the cases, so there are no pure shared-electron bonds in any of the alanates. There are, however, strong core attractors at the hydrogen positions for all the compounds, which is consistent with the polar covalent bond between Al and H. Accordingly, no localization is found near the Al cores. The alkali atoms all have spherical attractors around the cores, indicative of the ionic bonding between the alkali atoms and AlH4. It is necessary to include the 1 s state in the lithium potential to see the attractor around Li; the attractors around Na and K are both of p-type. TABLE II. Hirshfeld charges and overlap populations for the pure alanates as calculated by ADF-BAND. Two values are given when two non-equivalent positions of the atom give different results. Overlap population Hirshfeld charges (e) LiAlH4 NaAlH4 KAlH4 Li3AlH6 Na3AlH6 K3AlH6 M Al H M–H Al–H 0.18 0.27 0.35 0.16 0.24/0.26 0.30/0.34 0.33 0.25 0.21 0.24/0.27 0.16 0.11 −0.12/0.14 −0.13 −0.13/−0.14 −0.12 −0.15 −0.18 0.20 0.13 0.13 0.21 0.23 0.22 0.73 0.68 0.79 0.49 0.40 0.51 J. Mater. Res., Vol. 20, No. 12, Dec 2005 3205 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations reversible hydrogenation is relatively easily achieved in Na and K alanate but not in Li alanate, we need to point out significant and versatile differences between Li alanate and the others to make such predictions. Unfortunately, this does not seem to be possible, since the only significant difference that has been described here is the relatively high stability of the Li hexahydride, explained by the small size of the Li cation. Since this effect is specific to Li, it has a limited relevance to the prediction of other alanate phases. Complete thermodynamic calculations of the relevant phases thus probably need to be carried out to predict the temperature-dependent thermodynamic properties that designate potentially good hydrogen-storage materials. These calculations are more computationally demanding than mere ground-state calculations but yield more reliable predictions of hydrogenation properties. Efforts to move in this direction are in progress. B. Mixed-alkali alanates In the following, we present results for all possible combinations of the form MxM⬘(1−x)AlH4 (tetrahydrides) and MxM⬘(3−x)AlH6 (hexahydrides), where M and M⬘ are Li, Na, or K, and x can take the values 0, 0.25, 0.5, and 0.75 in the former case and 0, 0.5, 1, 1.5, 2, and 2.5 in the latter. This completes our search for mixed-alkali alanates. Their stability is defined relative to the monoalkali alanates, and may be measured by the mixing enthalpy Hmix Hmix[MxM⬘(n−x)AlH3+n] ⳱ E[MxM⬘(n−x)AlH3+n] – x E(MAlH3+n) – (n − x) E(M⬘AlH3+n) , FIG. 4. Electron localization function of the tetrahydrides LiAlH4, NaAlH4, and KAlH4. High values mean high degree of localization. The contour plots are taken through planes with all three atom types in or close to the plane, that is 013(LiAlH4), 100(NaAlH4), and 010(KAlH4). The approximate positions of in-plane or near-plane atoms are indicated in the plots. Unit cell directions are included in the bottom left corners. For the sake of comparison, the plots are out of scale. It would be very convenient to formulate certain properties from calculations to indicate whether or not a hydride is suitable as a hydrogen storage material, to expedite future searches of potential materials. Since 3206 (7) where n ⳱ 1 for the tetrahydrides and 3 for the hexahydrides. The results are presented in Table III and Fig. 5 for the tetrahydrides and in Table IV and Fig. 6 for the hexahydrides. No stable mixed tetrahydrides are found for any of the series with 25%, 50%, or 75% substitution. It is thus more difficult to dissolve a minor alkali species in the tetrahydrides. This is unfortunate, since mixed tetrahydrides would be needed to achieve a hydrogen storage system with higher hydrogen density than NaAlH4. Mixed-alkali alanate systems where stable hexahydrides exist are hence useful in the last reaction step [Eq. (2)] only. This has so far been reported only for LiNa2AlH6, and around 80% of the theoretical hydrogen content was obtained reversibly; that, is slightly less than 3 wt%.97,98 The reason it is difficult to form mixed-alkali tetrahydrides may to a certain extent be understood from the local coordination of the alkali atoms. While some of the alkali sites always remain in low coordination (octahedral) in the hexahydrides even when going to K alanate, the coordination of all alkali atoms change in the J. Mater. Res., Vol. 20, No. 12, Dec 2005 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations TABLE III. The most stable structures, densities, formation enthalpies, and mixing enthalpies of the mixed-alkali tetrahydrides, on the form MxM⬘(1−x)AlH4. The enthalpies are defined in the text. The end symmetry has been determined using an accuracy of 5 pm and 0.05°. The densities are based on the accessible hydrogen content (reactions 1 and 2.) H vol density (kg H/m3) MxM⬘(1−x) Most stable model This work Experimental This work Experimental H wt% Formation enthalpy (kJ/mol atom) Li Li0.75Na0.25 Li0.5Na0.5 Li0.25Na0.75 Na K0.25Na0.75 K0.5Na0.5 K0.75Na0.25 K K0.75Li0.25 K0.5Li0.5 K0.25Li0.75 LiAlD4 KAuCl4 KAuCl4 NaAlD4 NaAlD4 NaAlD4 NaAlD4 KAlD4 KAlD4 KAlD4 NaAlCl4 NaAlCl4 P21/c Pm Pc P4̄ I41/a P4̄ I4̄ Pm Pnma Pm P21 P1 P21/c ... ... ... I41/a ... ... ... Pnma ... ... ... 73.36 83.34 75.88 77.26 72.04 66.68 61.68 55.27 52.56 56.01 58.37 65.07 73.81 ... ... ... 72.64 ... ... ... 53.15 ... ... ... 7.97 7.21 6.58 6.05 5.60 5.21 4.87 4.58 4.31 4.87 5.60 6.57 −55.5 −53.9 −51.6 −50.8 −54.9 −54.6 −58.2 −64.1 −70.0 −61.8 −58.7 −53.6 Space group FIG. 5. Mixing enthalpy [defined in Eq. (7)] of the mixed-alkali tetrahydrides LixNa(1−x)AlH4 (solid curve), K(1−x)NaxAlH4 (dashed curve), and K(1−x)LixAlH4 (dotted curve) as a function of x. tetrahydrides (see Table I). The energy penalty for substituting a large atom with high coordination by a smaller atom is thus much larger than if the large atom has relatively low coordination in the first place; in the latter case there may actually be an energy gain, as we shall see for the hexahydrides. The mixed-alkali hexahydrides with two different alkali cations have already been studied elsewhere, and four stable compounds were identified.31,32 The only stoichiometries investigated in the previous studies were those in which x assumes integral values. When we extend the scope to intermediate stoichiometries, this represents ordered versions of solid solutions. These structures are created by using unit cells with Z at least Mixing enthalpy (kJ/mol atom) ... 0.5 1.2 1.4 ... 1.3 1.4 0.7 ... 1.5 1.3 1.8 equal to 2 (the conventional cell.) We also created supercells with Z ⳱ 4 to check different configurations of the two alkali atoms within the cell. In most of the cases, the difference in total energy between the different alternatives was negligible, and we have only used the configuration with largest separation distances between the minor alkali atoms. The three hexahydrides with a lighter alkali atom substituting one out of three alkali atoms of either Na3AlH6 or K 3 AlH 6 (that is, LiNa 2 AlH 6 , K 2 LiAlH 6 , or K2NaAlH6) are all stable compared to the pure alanates. In addition KNa2AlH6 is stable compared to the pure alanates but not when compared to K2NaAlH6 and Na3AlH6 (it is then unstable by 1.54 kJ/mol atom). This is consistent with the experimental results of Graetz et al., where only the three compounds with one lighter substitution atom were found.98 Our results are also consistent with their proposed crystal structures for the mixed compounds (all cubic in the Fm3̄m space group), except that of LiNa2AlH6, where we predict a monoclinic distortion. This is valid only at 0 K, however. There are signs from the lattice dynamics direct method that the cubic Fm3̄m structure may be more stable at higher temperatures.94 Moreover, a lower-symmetry phase of the same compound has been observed at low temperatures by powder neutron diffraction experiments.95 The reason these structures are stable may be understood from the local coordination of the different alkali atoms. In all the stable structures, the smallest alkali atom occupies an octahedrally coordinated site, while the larger atoms occupy sites with higher coordination (10–12). Since there are twice as many of the latter sites, it is not surprising that the preferred mixing factor is one third light alkali atoms. Results for the intermediary mixed alanates with actual stoichiometry MnM⬘(6−n)Al2H12 (n uneven) are also J. Mater. Res., Vol. 20, No. 12, Dec 2005 3207 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations TABLE IV. The most stable structures and mixing enthalpies of the mixed-alkali hexahydrides, on the form MxM⬘(3−x)AlH6. The hydrogen densities are the accessible amount when releasing hydrogen down to the alkali monohydrides. The mixing enthalpy is defined in the text. The end symmetry has been determined using an accuracy of 5 pm and 0.05°. MxM⬘(3−x) Most stable model This work Experiment This work Experiment H wt% Formation enthalpy (kJ/mol H2) Li3 Li2.5Na0.5 Li2Na Li1.5Na1.5 LiNa2 Li0.5Na2.5 Na3 K0.5Na2.5 KNa2 K1.5Na1.5 K2Na K2.5Na0.5 K3 K2.5Li0.5 K2Li K1.5Li1.5 KLi2 K0.5Li2.5 Ti3NiS6 Ti3NiS6 Na3AlD6 K3MoF6 Na3AlD6 Ti3NiS6 Na3AlD6 Li3AlF6 Na3AlD6 K3MoF6 Na3AlD6 K3MoF6 Na3AlD6 K3Fe(CN)6 Na3AlD6 K3MoF6 Na3AlD6 Ti3NiS6 R3̄ P1 P21 P1 P21/n P1 P21/n P21 P21 P4̄m2 Fm3̄m P4/mmm P21/n P1 Fm3̄m P4̄m2 P21 P1 R3̄ ... ... ... Fm3̄m ... P21/n ... ... ... Fm3̄m ... ... ... Fm3̄m ... ... ... 57.20 53.36 54.94 52.44 50.46 42.58 43.85 41.09 40.08 38.96 37.85 34.81 31.83 33.05 41.87 44.75 47.03 46.45 56.83 ... ... ... 49.46 ... 43.74 ... ... ... 37.50 ... ... ... 40.15 ... ... ... 5.62 4.89 4.33 3.88 3.52 3.22 2.96 2.75 2.56 2.40 2.56 2.13 2.01 2.25 2.56 2.96 3.52 4.32 −102.8 −94.0 −88.2 −85.4 −84.5 −71.7 −69.9 −68.9 −75.8 −82.8 −92.1 −85.3 −78.5 −77.8 −100.5 −89.5 −87.7 −90.2 H vol density (kg H/m3) Space group FIG. 6. Mixing enthalpy (defined in the text) of the mixed-alkali hexahydrides LixNa(3−x)AlH6 (solid curve), K(3−x)NaxAlH6 (dashed curve), and K(3−x)LixAlH6 (dotted curve) as a function of x. included in Table IV. The only stable compound in this form is K2.5Na0.5AlH6. This is also a cubic structure in the Fm3̄m space group, and it is marginally stable compared to K2NaAlH6 and K3AlH6. This suggests that Na is soluble in K3AlH6 from low levels and up to at least 33% substitution of K by Na. Since the crystal structure is the same for all three levels of substitution (0%, 17%, and 33%), the only way this may be detected by diffraction experiments is probably by the changed lattice parameters. They are shown indirectly in Table III (the volume 3208 Mixing enthalpy (kJ/mol atom) ... 1.0 1.1 0.3 −1.1 1.1 ... 0.7 −0.9 −2.6 −4.9 −2.5 ... 1.4 −4.2 0.3 2.1 2.6 may be found from the hydrogen volume density) and are a ⳱ 810, 832, and 858 pm for K 2 NaAlH 6 , K 2.5 Na 0.5 AlH 6 , and K 3 AlH 6 . The latter compound deviates slightly from cubic symmetry, and the lattice constants for K3AlH6 in the cubic-like supercell are 852, 858, and 863 pm. It seems that the lattice constant increases linearly with increasing Na content and that the changes are large enough to be seen in diffraction experiments. The compound with 50% Na and K is also stable compared to the pure alanates, but it is unstable compared to K2NaAlH6 and Na3AlH6. The solubility limit for Na in K3AlH6 is thus between 33% and 50%, probably near 33% if judged from the curve. We have not found any stable intermediate compounds between K2LiAlH6 and K3AlH6 or between LiNa2AlH6 and Na3AlH6. This may be because we have not found the correct crystal structure for these structures. In that case, however, the crystal structure must be significantly different in the intermediate compound, and we no longer speak of a solid solution, but rather a distinct phase. This leads us to propose that the solubilities of Li in K3AlH6 and Na3AlH6 do not represent a general trend, and that only the ordered phases with 33% Li exist. We have used the same configuration of alkali atoms of the phase Na1.5Li1.5AlH6 as described in Ref. 81, but while they obtained a stable structure (by approximately 0.4 kJ/mol atom), ours is slightly unstable compared to the pure alanates (by approximately 0.3 kJ/mol atom). The difference is not very big, however, and is clearly within what could be expected when comparing results from different methods (they used, for J. Mater. Res., Vol. 20, No. 12, Dec 2005 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations instance, ultrasoft pseudopotentials in Ref. 81, while we use the projector augmented wave method.) In any case, Na1.5Li1.5AlH6 is unstable compared to LiNa2AlH6 and Na3AlH6, so we do not expect that it will be observed in experiments, unless it exists as a metastable phase. In this study, we have investigated a rather large number of hypothetical mixed-alkali structures, but the field is not necessarily exhausted. Since we have included only a limited number of input models in the relaxation procedure, there is a chance that some compounds have more stable structures not accessible from our set of input structures. Low-symmetry structures may be made accessible by deliberately breaking symmetry of highersymmetry structures,111 but this makes the calculations much more expensive. It is also possible to investigate a larger number of different configurations of the alkali atoms; we have only done this in the most stable structures of the hexahydrides, based on the Na3AlD6 and Li3AlD6 structures. Again, if a large number of different configurations were included for all the input structures of all the hypothetical compounds, the computational effort needed would be insurmountable with our current resources. Since we have calculated only ordered structures in this study, we also expect that configurational entropy may stabilize some of the mixed compounds. Most importantly, temperature effects included by for instance phonon calculations would give valuable information on the relative stability of the various phases as a function of temperature and pressure. Nevertheless, there is small chance that any of these methods provide stabilization effects sufficient for some of the light mixed-alkali tetrahydrides to be found stable compared to the pure ones. This means that NaAlH4 is the lightest remaining alanate with proven reversible hydrogenation at moderate conditions and that mixing of the LiAlH4 and NaAlH4 phases does not seem to be a feasible way toward a viable hydrogen storage system. IV. SUMMARY AND OUTLOOK The crystal structure and detailed electronic structure of pure and mixed alkali alanates in the ground state have been calculated by density-functional calculations within the generalized gradient approximation. The correspondence with experimentally known values is excellent in almost all the cases. All known monoalkali crystal structures are reproduced within very small deviations, and the crystal structure of K3AlH6 has been predicted. The only exception to this very good correspondence is LiNa2AlH6, where the predicted crystal structure is weakly monoclinic, whereas the experimental one is cubic. This discrepancy may, however, be entirely due to the fact that our models give equilibrium structures at a temperature of 0 K while experimental work is performed at higher temperatures A recent diffraction study of this compound95 reveals a low-temperature phase of this compound with lower symmetry that may be consistent with our proposed crystal structure at 0 K. It is important to know why it is much more difficult to hydrogenate Li alanate from the gas phase than the heavier compounds, since Li alanate would have been a highly desirable hydrogen-storage material from a gravimetric point of view. One reason seems to be the relatively high stability of the Li3AlH6 phase, which has a totally different crystal structure than the heavier hexahydrides, apparently because of the small size of the Li cation. Since no similar, more stable structure exists for LiAlH4, the relative stability of the tetra- and hexahydrides is different in Li alanate from that in the Na and K alanates, and thus hydrogenation of Li alanate is more difficult. It is also likely that kinetic barriers hinder the hydrogenation of Li alanates, particularly of the hexahydrides [reverse reaction (2)]. The electronic structures of the different alanates are all quite similar. They reveal a polar covalent bond within the aluminohydride complexes and ionic bonding between the alkali cation and the complex. Lithium shows a stronger degree of covalence in its interaction with hydrogen than the heavier alkali metals. This may be readily explained by its low polarizability and high ionization potential compared to the latter. The alanates are insulators, with calculated band gaps around 5 eV for the tetrahydrides and 2.5–3 eV for the hexahydrides. These band gaps are probably underestimated by 2–3 eV. The presently reported work completes our search for mixed-alkali alanates, and no stable mixed-alkali tetrahydride phases [on the form MxM⬘(1−x)AlH4 with M and M⬘ ⳱ Li, Na, K, and x ⳱ 0.25,0.50, and 0.75] were found. The only mixed-alkali hexahydride phases [in the form MxM⬘(3−x)AlH6 with M and M⬘ ⳱ Li, Na, K, and x ⳱ 0.5, 1, 1.5, 2, and 2.5] found to be stable were LiNa2AlH6, K2LiAlH6, K2NaAlH6, and K2.5Na0.5AlH6. In addition, K1.5Na1.5AlH6 and KNa2AlH6 were found to be stable compared to the pure alanates, but unstable compared to K2NaAlH6 and the pure alanates at the ground state. The shape of the calculated mixing enthalpy as a function of the M⬘ content suggests that there is general solubility of Na in K3AlH6 up to 1/3 Na, but the LiNa2AlH6 and K2LiAlH6 phases are unique structures with possibly highly ordered distribution of the alkali atoms. We see from Tables I, III, and IV that the only (existing) alkali alanates that satisfy the international density targets for hydrogen-storage materials112 of 5 wt% and 70 kg/m3 are LiAlH4 and NaAlH4. If the volumetric density is not too important, Li3AlH6 could also be an J. Mater. Res., Vol. 20, No. 12, Dec 2005 3209 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations interesting material, with the same gravimetric hydrogen density as NaAlH4 released in a single step. Since reversible hydrogenation of Li alanates does not occur at reasonable conditions,35 however, the only remaining usable alkali alanate satisfying this goal is NaAlH4. Further, if the United States Department of Energy Freedom CAR target of 6 wt% hydrogen for the storage system113 is to be followed, only LiAlH4 has a sufficiently high hydrogen density among the alkali alanates. Hence, the Freedom CAR density target will probably not be reached using alkali alanates. Even the international target of 5 wt% may be difficult to reach using NaAlH4, since only partial hydrogenation is achievable with reasonable kinetics at moderate hydrogen pressure (which is important to avoid the need for bulky containers for the hydride). This means that one has to search among other groups of hydrides for lighter storage materials if the international goals are to be reached. Possible candidates may include alkaline earth alanates, borohydrides, imides and amides, or magnesium-based hydrides. The ideal system has not yet been found, and the search continues. Hydrogen storage is a crucial, still unresolved part of a future hydrogen energy system, and the feasibility of such a system may depend strongly upon the success of this search. ACKNOWLEDGMENTS Economic support from the NANOMAT program of the Norwegian research council and supercomputing resources from the NOTUR project are acknowledged by OML and OS. Economic support by the United States Department of Energy under Contract No. DE-FC36014012 is acknowledged by SMO. REFERENCES 1. A. Züttel: Materials for hydrogen storage. Mater. Today 6(9) 24 (2003). 2. G. Sandrock and R.C. Bowman, Jr.: Gas-based hydride applications: Recent progress and future needs. J. Alloys Compd. 356–357, 794 (2003). 3. A. Züttel: Hydrogen storage methods. Naturwissenschaften 91, 157 (2003). 4. A.M. Seayad and D.M. Antonelli: Recent advances in hydrogen storage in metal-containing inorganic nanostructures and related materials. Adv. Mater. 16, 765 (2004). 5. P.P.E.W. Grochala: Thermal decomposition of the noninterstitial hydrides for the storage and production of hydrogen. Chem. Rev. 104, 1283 (2004). 6. F. Schüth, B. Bogdanovic, and M. Felderhoff: Light metal hydrides and complex hydrides for hydrogen storage. Chem. Commun. 2249 (2004). 7. M. Conte, P.P. Prosini, and S. Passerini: Overview of energy/ hydrogen storage: State-of-the-art of the technologies and prospects for nanomaterials. Mater. Sci. Eng. B (in press). 3210 8. B. Bogdanovic and M. Schwickardi: Ti-doped alkali metal aluminium hydrides as potential novel reversible hydrogen-storage materials. J. Alloys Compd. 253, 1 (1997). 9. C.M. Jensen, R. Zidan, N. Mariels, A. Hee, and C. Hagen: Advanced titanium doping of sodium aluminum hydride: Segue to a practical hydrogen storage material? Int. J. Hydrogen Energy 24, 461 (1999). 10. A. Zaluska, L. Zaluski, and J.O. Ström-Olsen: Sodium alanates for reversible hydrogen storage. J. Alloys Compd. 298, 125 (2000). 11. K.J. Gross, G.J. Thomas, and C.M. Jensen: Catalyzed alanates for hydrogen storage. J. Alloy Compd. 330–332, 683 (2002). 12. G.P. Meisner, G.G. Tibbetts, F.E. Pinkerton, C.H. Olk, and M.P. Balogh: Enhancing low pressure hydrogen storage in sodium alanates. J. Alloys Compd. 337, 254 (2002). 13. M. Fichtner, O. Fuhr, O. Kircher, and J. Rothe: Small Ti clusters for catalysis of hydrogen exchange in NaAlH4. Nanotechnol. 14, 778 (2003). 14. T.N. Dymova, D.P. Aleksandrov, V.N. Konoplev, T.A. Silina, and A.S. Sizareva: Spontaneous and thermal decomposition of lithium tetrahydridoaluminate LiAlH4: The promoting effect of mechanochemical action on the process. Russ. J. Coord. Chem. 20, 263 (1994). 15. T.N. Dymova, V.N. Konoplev, D.P. Aleksandrov, A.S. Sizareva, and T.A. Silina: A novel view of the nature of chemical- and phase-composition modifications in lithium hydridoaluminates LiAlH4 and Li3AlH6 on heating. Russ. J. Coord. Chem. 21, 165 (1995). 16. N.N. Mal’tseva, A.I. Golovanova, T.N. Dymova, and D.P. Aleksandrov: Russ. J. Inorg. Chem. 46, 1793 (2001). 17. V.P. Balema, K.W. Dennis, and V.K. Pecharsky: Rapid solidstate transformation of tetrahedral [AlH4]− into octahedral [AlH6]3− in lithium aluminohydride. Chem. Commun. 1665 (2000). 18. V.P. Balema, J.W. Wiench, K.W. Dennis, M. Pruski, and V.K. Pecharsky: Titanium catalyzed solid-state transformations in LiAlH4 during high-energy ball milling. J. Alloys Compd. 329, 108 (2001). 19. J. Chen, N. Kuriyama, Q. Xu, H.T. Takeshita, and T. Sakai: Reversible hydrogen storage via titanium-catalyzed LiAlH4 and Li3AlH6. J. Phys. Chem. B 105, 11214 (2001). 20. B.C. Hauback, H.W. Brinks, and H. Fjellvåg: Accurate structure of LiAlD4 studied by combined powder neutron and x-ray diffraction. J. Alloys Compd. 346, 184 (2002). 21. H.W. Brinks and B.C. Hauback: The structure of Li3AlD6. J. Alloys Compd. 354, 143 (2003). 22. H.W. Brinks, B.C. Hauback, P. Norby, and H. Fjellvåg: The decomposition of LiAlD4 studied by in-situ x-ray and neutron diffraction. J. Alloys Compd. 351, 222 (2003). 23. P. Vajeeston, P. Ravindran, R. Vidya, H. Fjellvåg, and A. Kjekshus: Huge-pressure-induced volume collapse in LiAlH4 and its implications to hydrogen storage. Phys. Rev. B 68, 212101 (2003). 24. P. Vajeeston, P. Ravindran, R. Vidya, H. Fjellvåg, and A. Kjekshus: Pressure-induced phase of NaAlH4: A potential candidate for hydrogen storage? Appl. Phys. Lett. 82, 2257 (2003). 25. O.M. Løvvik: Periodic band calculation on low index surfaces of crystalline LiAlH4. J. Alloys Compd. 356–357, 178 (2003). 26. S-C. Chung and H. Morioka: Thermochemistry and crystal structures of lithium, sodium and potassium alanates as determined by ab initio simulations. J. Alloys. Compd. 372, 92 (2004). 27. D. Blanchard, H.W. Brinks, B.C. Hauback, and P. Norby: Desorption of LiAlH4 with Ti- and V-based additives. Mater. Sci. Eng. B-Solid State Mater. Adv. Technol. 108, 54 (2004). J. Mater. Res., Vol. 20, No. 12, Dec 2005 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations 28. P. Vajeeston, P. Ravindran, A. Kjekshus, and H. Fjellvåg: Structural stability and electronic structure for Li3AlH6. Phys. Rev. B 69, 020104 (2004). 29. O.M. Løvvik, S.M. Opalka, H.W. Brinks, and B.C. Hauback: Crystal structure and thermodynamic stability of the lithium alanates LiAlH4 and Li3AlH6. Phys. Rev. B 69, 134117 (2004). 30. O.M. Løvvik: Adsorption of Ti on LiAlH4 surfaces studied by band-structure calculations. J. Alloys Compd. 373, 28 (2004). 31. O.M. Løvvik and O. Swang: Structure and stability of possible new alanates. Europhys. Lett. 607, 607 (2004). 32. O.M. Løvvik and O. Swang: Crystal structures and electronic structures of alkali aluminohexahydrides from density-functional calculations. J. Alloys Compd. (in press). 33. J.K. Kang, J.Y. Lee, R.P. Muller, and W.A. Goddard: Hydrogen storage in LiAlH4: Predictions of the crystal structures and reaction mechanisms of intermediate phases from quantum mechanics. J. Chem. Phys. 121, 10623 (2004). 34. C.M. Andrei, J.C. Walmsley, H.W. Brinks, R. Holmestad, S.S. Srinivasan, C.M. Jensen, and B.C. Hauback: Electron microscopy studies of lithium aluminium hydrides. J. Alloys Compd. (in press). 35. H.W. Brinks, A. Fossdal, J.E. Fonneløp, and B.C. Hauback: Crystal structure and stability of LiAlH4 with TiF3 additive. J. Alloys Compd. (in press). 36. V.P. Balema, V.K. Pecharsky, and K.W. Dennis: Solid-state phase transformations in LiAlH4 during high-energy ball milling. J. Alloys Compd. 313, 69 (2000). 37. R.A. Zidan, S. Takara, A.G. Hee, and C.M. Jensen: Hydrogen cycling behavior of zirconium and titanium-zirconium-doped sodium aluminum hydride. J. Alloys Compd. 285, 119 (1999). 38. B. Bogdanovic, R.A. Brand, A. Marjanovic, M. Schwickardi, and J. Tolle: Metal-doped sodium aluminium hydrides as potential new hydrogen-storage materials. J. Alloys Compd. 302, 36 (2000). 39. K.J. Gross, S. Guthrie, S. Takara, and G.J. Thomas: In-situ x-ray diffraction study of the decomposition of NaAlH4. J. Alloys Compd. 297, 270 (2000). 40. A. Zaluska, L. Zaluski, and J.O. Ström-Olsen: Structure, catalysis and atomic reactions on the atomic scale: A systematic approach to metal hydrides for hydrogen storage. Appl. Phys. A 72, 157 (2001). 41. C.M. Jensen and K.J. Gross: Development of catalytically enhanced sodium aluminum hydride as a hydrogen-storage material. Appl. Phys. A 72, 213 (2001). 42. B. Bogdanovic and M. Schwickardi: Ti-doped NaAlH4 as a hydrogen-storage material–preparation by Ti-catalyzed hydrogenation of aluminum powder in conjunction with sodium hydride. Appl. Phys. A 72, 221 (2001). 43. G.J. Thomas, K.J. Gross, N.Y.C. Yang, and C.M. Jensen: Microstructural characterization of catalyzed NaAlH4. J. Alloys Compd. 330–332, 702 (2002). 44. G. Sandrock, K.J. Gross, G.J.T.C. Jensen, D. Meeker, and S. Takara: Engineering considerations in the use of catalyzed sodium alanates for hydrogen storage. J. Alloys Compd. 330–332, 696 (2002). 45. D. Sun, T. Kiyobayashi, H.T. Takeshita, N. Kuriyama, and C.M. Jensen: X-ray diffraction studies of titanium and zirconium doped NaAlH4: Elucidation of doping induced structural changes and their relationship to enhanced hydrogen storage properties. J. Alloys Compd. 337, L8 (2002). 46. K.J. Gross, G. Sandrock, and G.J. Thomas: Dynamic in-situ xray diffraction of catalyzed alanates. J. Alloys Compd. 330–332, 691 (2002). 47. D.L. Anton: Hydrogen desorption kinetics in transition metal modified NaAlH4. J. Alloys Compd. 356–357, 400 (2003). 48. K.J. Gross, E.H. Majzoub, and S.W. Spangler: The effects of titanium precursors on hydriding properties of alanates. J. Alloys Compd. 356–357, 423 (2003). 49. M.P. Balogh, G.G. Tibbetts, F.E. Pinkerton, G.P. Meisner, and C.H. Olk: Phase changes and hyrogen release during decomposition of sodium alanates. J. Alloys Compd. 350, 136 (2003). 50. T. Kiyobayashi, S.S. Srinivasan, D. Sun, and C.M. Jensen: Kinetic study and determination of the enthalpies of activation of the dehydrogenation of titanium- and zirconium-doped NaAlH4 and Na3AlH6. J. Phys. Chem. A 107, 7671 (2003). 51. D. Sun, S.S. Srinivasan, T. Kiyobayashi, N. Kuriyama, and C.M. Jensen: Rehydrogenation of dehydrogenated NaAlH4 at low temperature and pressure. J. Phys. Chem. B 107, 10176 (2003). 52. B. Bogdanovic, M. Felderhoff, M. Germann, M. Härtel, A. Pommerin, F. Schüth, C. Weidenthaler, and B. Zibrowius: Investigation of hydrogen discharging and recharging processes of Ti-doped NaAlH4 by x-ray diffraction analysis (XRD) and solid-state NMR spectroscopy. J. Alloys Compd. 350, 246 (2003). 53. B. Bogdanovic, M. Felderhoff, S. Kaskel, A. Pommerin, K. Schlichte, and F. Schüth: Improved hydrogen storage properties of Ti-doped sodium alanate using titanium nanoparticles as doping agents. Adv. Mater. 15, 1012 (2003). 54. E.H. Majzoub and K.J. Gross: Titanium-halide catalystprecursors in sodium aluminum hydrides. J. Alloys Compd. 356–357, 363 (2003). 55. C. Weidenthaler, A. Pommerin, M. Felderhoff, B. Bogdanovic, and F. Schüth: On the state of the titanium and zirconium in Tior Zr-doped NaAlH4 hydrogen storage material. Phys. Chem. Chem. Phys. 5, 5149 (2003). 56. M. Fichtner, J. Engel, O. Fuhr, O. Kircher, and O. Rubner: Nanocrystalline aluminium hydrides for hydrogen storage. Mater. Sci. Eng. B 108, 42 (2004). 57. O. Kircher and M. Fichtner: Hydrogen exchange kinetics in NaAlH4 catalyzed in different decomposition states. J. Appl. Phys. 95, 7748 (2004). 58. S.S. Srinivasan, H.W. Brinks, B.C. Hauback, D. Sun, and C.M. Jensen: Long term cycling behavior of titanium doped NaAlH4 prepared through solvent mediated milling of NaH and Al with titanium dopant precursors. J. Alloys Compd. 377, 283 (2004). 59. W. Luo and K.J. Gross: A kinetics model of hydrogen absorption and desorption in Ti-doped NaAlH4. J. Alloys Compd. 385, 224 (2004). 60. A. Léon, O. Kircher, J. Rothe, and M. Fichtner: Chemical state and local structure around titanium atoms in NaAlH4 doped with TiCl3 using x-ray absorption spectroscopy. J. Phys. Chem. B 108, 16372 (2004). 61. V. Ozolins, E.H. Majzoub, and T.J. Udovic: Electronic structure and Rietveld refinement parameters of Ti-doped sodium alanates. J. Alloys Compd. 375, 1 (2004). 62. M. Felderhoff, K. Klementiev, W. Grünert, B. Spliethoff, B. Tesche, J.M.B. von Colbe, B. Bogdanovic, M. Härtel, A. Pommerin, F. Schüth, and C. Weidenthaler: Combined TEM-EDX and XAFS studies of Ti-doped sodium alanate. Phys. Chem. Chem. Phys. 6, 4369 (2004). 63. H.W. Brinks, C.M. Jensen, S.S. Srinivasan, B.C. Hauback, D. Blanchard, and K. Murphy: Synchrotron x-ray and neutron diffraction studies of NaAlH4 containing Ti additives. J. Alloys Compd. 376, 215 (2004). 64. D. Sun, S.S. Srinivasan, G. Chen, and C.M. Jensen: Rehydrogenation and cycling studies of dehydrogenated NaAlH4. J. Alloys Compd. 373, 265 (2004). J. Mater. Res., Vol. 20, No. 12, Dec 2005 3211 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations 65. R.T. Walters and J.H. Scogin: A reversible hydrogen storage mechanism for sodium alanate: The role of alanes and the catalytic effect of the dopant. J. Alloys Compd. 379, 135 (2004). 66. P. Wang and C.M. Jensen: Method for preparing Ti-doped NaAlH4 using Ti powder: Observation of an unusual reversible dehydrogenation behavior. J. Alloys Compd. 379, 99 (2004). 67. P. Wang and C.M. Jensen: Preparation of Ti-doped sodium aluminum hydride from mechanical milling of NaH/Al with off-theshelf Ti powder. J. Phys. Chem. B 108, 15827 (2004). 68. J. Graetz, J.J. Reilly, J. Johnson, A.Y. Ignatov, and T.Y. Tyson: X-ray absorption study of Ti-activated sodium aluminum hydride. Appl. Phys. Lett. 85, 500 (2004). 69. J. Iniguez, T. Yildirim, T.J. Udovic, M. Sulic, and C.M. Jensen: Structure and hydrogen dynamics of pure and Ti-doped sodium alanate. Phys. Rev. B 70, 060101 (2004). 70. O.M. Løvvik and S.M. Opalka: Density-functional calculations of Ti-enhanced NaAlH4. Phys. Rev. B 71, 054103 (2005). 71. J. Wang, A.D. Ebner, T. Prozorov, R. Zidan, and J.A. Ritter: Effect of graphite as co-dopant on the dehydrogenation and hydrogenation kinetics of Ti-doped sodium aluminum hydride. J. Alloys Compd. (in press). 72. J. Wang, A.D. Ebner, R. Zidan, and J.A. Ritter: Synergistic effects of co-dopants on the dehydrogenation kinetics of sodium aluminum hydride. J. Alloys Compd. 391, 245 (2005). 73. S. Gomes, G. Renaudin, H. Hagemann, K. Yvon, M.P. Sulic, and C.M. Jensen: Effects of milling, doping and cycling of NaAlH4 studied by vibrational spectroscopy and x-ray diffraction. J. Alloys Compd 390, 305 (2005). 74. A.G. Haiduc, H.A. Stil, M.A. Schwarz, P. Paulus, and J.J.C. Geerlings: On the fate of the Ti catalyst during hydrogen cycling of sodium alanate. J. Alloys Compd. 393, 252 (2005). 75. E.H. Majzoub, J.L. Herberg, R. Stumpf, S. Spangler, and R.S. Maxwell: XRD and NMR investigation of Ti-compound formation in solution-doping of sodium aluminum hydrides: Solubility of Ti in NaAlH4 crystals grown in THF. J. Alloys Compd. (in press). 76. M. Resana, M.D. Hamptona, J.K. Lomnessa, and D.K. Slattery: Effect of TixAly catalysts on hydrogen storage properties of LiAlH4 and NaAlH4. Int. J. Hydrogen Energy (in press). 77. C.M. Andrei, J.C. Walmsley, H.W. Brinks, R. Holmestad, S.S. Srinivasan, C.M. Jensen, and B.C. Hauback: Electronmicroscopy studies of NaAlH4 with TiF3 additive: Hydrogencycling effects. Appl. Phys. A 80, 709 (2005). 78. S.M. Opalka and D.L. Anton: First principles study of sodiumaluminum-hydrogen phases. J. Alloys Comp. 356–357, 486 (2003). 79. A. Aguayo and D.J. Singh: Electronic structure of the complex hydride NaAlH4. Phys. Rev. B 69, 155103 (2004). 80. D.J. Ross, M.D. Halls, A.G. Nazro, and R.F. Aroca: Raman scattering of complex sodium aluminimum hydride for hydogen storage. Chem. Phys. Lett. 388, 430 (2004). 81. M.E. Arroyo y de Dompablo and G. Ceder: First principles investigations of complex hydrides AMH4 and A3MH6 (A ⳱ Li, Na, K, M⳱B, Al, Ga) as hydrogen storage systems. J. Alloys Compd. 364, 6 (2004). 82. A. Peles, J.A. Alford, Z. Ma, L. Yang, and M.Y. Chou: Firstprinciples study of NaAlH4 and Na3AlH6 complex hydrides. Phys. Rev. B 70, 165105 (2004). 83. X. Ke and I. Tanaka: Decomposition reactions for NaAlH4, Na3AlH6, and NaH: First-principles study. Phys. Rev. B 71, 024117 (2005). 84. E.H. Majzoub, K.F. McCarty, and V. Ozolins: Lattice dynamics 3212 of NaAlH 4 from high-temperature single-crystal Raman scattering and ab initio calculations: Evidence of highly stable AlH4− anions. Phys. Rev. B 71, 024118 (2005). 85. L. Hedin: New method for calculating 1-particle Greens function with application to electron-gas problem. Phys. Rev. 139, A796 (1965). 86. M.S. Hybertsen and S.G. Louie: Electron correlation in semiconductors and insulators: Band-gaps and quasi-particle energies. Phys. Rev. B 34, 5390 (1986). 87. J.P. Bastide, B. Bonnetot, J-M. Letoffe, and P. Claudy: Polymorphic transition of the trisodium hexahydroaluminate Na3AlH6. Mater. Res. Bull. 16, 91 (1981). 88. H. Morioka, K. Kakizaki, S.C. Chung, and A. Yamada: Reversible hydrogen decomposition of KAlH4. J. Alloys Compd. 353, 310 (2003). 89. B.C. Hauback, H.W. Brinks, R.H. Heyn, R. Blom, and H. Fjellvåg: The crystal structure of KAlD4. J. Alloys Comps. (in press). 90. P. Vajeeston, P. Ravindran, A. Kjekshus, and H. Fjellvåg: Crystal structure of KAlH4 from first principle calculations. J. Alloys Compd. 363, L7 (2004). 91. P. Claudy, B. Bonnetot, J-P. Bastide, and J-M. Létoffé: Reactions of lithium and sodium aluminium hydride with sodium or lithium hydride. Preparation of a new alumino-hydrode of lithium and sodium LiNa2AlH6. Mater. Res. Bull. 17, 1499 (1982). 92. L. Zaluski, A. Zaluska, and J.O. Ström-Olsen: Hydrogenation properties of complex alkali metal hydrides fabricated by mechano-chemical synthesis. J. Alloys Compd. 290, 71 (1999). 93. J. Huot, S. Boily, V. Guther, and R. Schulz: Synthesis of Na3AlH6 and Na2LiAlH6 by mechanical alloying. J. Alloys Compd. 283, 304 (1999). 94. S. M. Opalka, P. Saxe, and O. M. Løvvik: Phonon calculations on the mixed-alkali phase LiNa2AlH6. (unpublished). 95. H.W. Brinks (private communication). 96. R. Genma, H.H. Uchida, N. Okada, and Y. Nishi: Hydrogen reactivity of Li-containing hydrogen-storage materials. J. Alloys Compd. 356, 358 (2003). 97. N. Okada, R. Genma, Y. Nishi, and H.H. Uchida: RE-oxide doped alkaline hydrogen-storage materials prepared by mechanical activation. J. Mater. Sci. 39, 5503 (2004). 98. J. Graetz, Y. Lee, J. J. Reilly, S. Park, and T. Vogt: Structure and thermodynamics of the mixed alkali alanates. (unpublished). 99. J.P. Bastide, P. Claudy, J.M. Letoffe, and J. El Hajri: Preparation and characterization of KAlH4. Rev. Chim. Miner. 24, 248 (1987). 100. G. Kresse and J. Hafner: Ab initio molecular-dynamics for liquid-metals. Phys. Rev. B 47, R558 (1993). 101. G. Kresse and J. Furthmüller: Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169 (1996). 102. G. Kresse and D. Joubert: From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 59, 1758 (1999). 103. J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh, and C. Fiolhais: Atoms, molecules, solids, and surfaces: Applications of the generalized gradient approximation for exchange and correlation. Phys. Rev. B 46, 6671 (1992). 104. G. te Velde and E.J. Baerends: Precise density-functional method for periodic structures. Phys. Rev. B 44, 7888 (1991). 105. G. te Velde and E.J. Baerends: Numerical integration of polyatomic systems. J. Comput. Phys. 99, 84 (1992). 106. E. Rönnebro, D. Noreus, K. Kadir, A. Reiser, and B. Bogdanovic: J. Mater. Res., Vol. 20, No. 12, Dec 2005 O.M. Løvvik et al.: Modeling alkali alanates for hydrogen storage by density-functional band-structure calculations 107. 108. 109. 110. Investigation of the perovskite related structures of NaMgH3, NaMgF, and Na3AlH6. J. Alloys Compd. 299, 101 (2000). B.C. Hauback, H.W. Brinks, C.M. Jensen, K. Murphy, and A.J. Maeland: Neutron diffraction structure determination of NaAlD4. J. Alloys Compd. 358, 142 (2003). R.D. Shannon: Revised effective ionic-radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Crystallogr. A 32, 751 (1976). WebElements Periodic table. http://www.webelements.com. B. Silvi and A. Savin: Classification of chemical-bonds based on topological analysis of electron localization functions. Nature 371, 683 (1994). 111. O.M. Løvvik: Predicted crystal structure of calcium alanate Ca(AlH4)2 from density-functional band-structure calculations. Phys. Rev. B (in press). 112. International energy agency hydrogen program, task 17. http:// www.ieahia.org/tasks/task17.html 113. The United States Department of Energy Freedom CAR targets. http://www.hydrogen.energy.gov/. J. Mater. Res., Vol. 20, No. 12, Dec 2005 3213