Large-Scale Screening of Metal Hydride Mixtures for High-Capacity

Mar 18, 2008 - Department of Chemical Engineering, Carnegie Mellon University, ... University of Pittsburgh, Pittsburgh, Pennsylvania 15261, National ...
0 downloads 0 Views 146KB Size
Download PDF
5258

2008, 112, 5258-5262 Published on Web 03/18/2008

Large-Scale Screening of Metal Hydride Mixtures for High-Capacity Hydrogen Storage from First-Principles Calculations Sudhakar V. Alapati,† J. Karl Johnson,‡,§ and David S. Sholl*,†,§,| Department of Chemical Engineering, Carnegie Mellon UniVersity, Pittsburgh, PennsylVania 15213, Department of Chemical and Petroleum Engineering, UniVersity of Pittsburgh, Pittsburgh, PennsylVania 15261, National Energy Technology Laboratory, Pittsburgh, PennsylVania 15236, and School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332 ReceiVed: January 22, 2008; In Final Form: February 25, 2008

First-principles calculations have been used to systematically screen >16 million mixtures of metal hydrides and related compounds to find materials with high capacity and favorable thermodynamics for reversible storage of hydrogen. These calculations are based on a library of crystal structures of >200 solid compounds comprised of Al, B, C, Ca, K, Li, Mg, N, Na, Sc, Si, Ti, V, and H. Thermodynamic calculations are used that rigorously describe the reactions available within this large library of compounds. Our calculations extend previous efforts to screen mixtures of this kind by orders of magnitude and, more importantly, identify multiple reactions that are predicted to have favorable properties for reversible hydrogen storage.

Development of practical methods to reversibly store large quantities of H2 is a serious hurdle facing the widespread use of H2 as a transportation fuel. The requirements for on-board H2 storage for a fuel cell vehicle to have a performance similar to those of current automobiles are well known.1,2 Long-term system targets include reversible H2 discharge of >9 wt % and >81 kg H2/m3 at moderate pressures and temperatures with rapid charging and discharging kinetics, high H2 purity, acceptable cost, and long operational life.3 Despite vigorous efforts, no material is known that achieves all of these goals.4-6 Light metal hydrides are well known for having a high capacity for H2 storage. Two prototypical examples are MgH2 (7.7 wt % H, 108.6 kg H2/m3) and LiBH4 (13.9 wt % H, 92.7 kg H2/m3when decomposing to LiH + B + 1.5H2). These examples are also characteristic of metal hydrides in that their decomposition temperatures are much higher than what is desirable for fuel cell applications. A convenient way to characterize the decomposition temperature is using the conditions where the metal hydride is in equilibrium with gaseous H2 at Peq ) 1 bar. With this criterion, MgH2 (LiBH4) decomposes at Teq ≈ 570 (685) K.7,8 An ideal high-capacity metal hydride for practical use would decompose with T between 273 and 373 K (0-100 °C). A powerful way to change the Teq of metal hydrides is to use mixtures of compounds that form reaction products different from those of the direct decomposition reactions. Vajo et al. showed that 2LiBH4 + MgH2 reacts reversibly forming MgB2 + 2LiH + 4H2 with H2 release of >9 wt % H and Teq ≈ 545 K.7 Another example of this concept is the work of Chen et al. * Corresponding author. E-mail: [email protected]. † Carnegie Mellon University. ‡ University of Pittsburgh. § National Energy Technology Laboratory. | Georgia Institute of Technology.

10.1021/jp800630s CCC: $40.75

on LiH + LiNH2.9 Although the Teq for these mixtures are more favorable than their parent hydrides, they are still higher than what is ideally desirable. For metal hydride mixtures to reach the storage targets outlined above, two independent challenges must be overcome. First, the enormous number of possible mixtures must be screened to find the (likely small) collection of mixtures with favorable storage and thermodynamic properties. Second, the reaction kinetics of these mixtures must be controlled. This second task can, in principle, be approached using appropriate catalysts and related techniques.10-12 Thermodynamic constraints, however, must be satisfied before considering kinetic issues. In previous work, we used first-principles calculations to predict the decomposition thermodynamics of >200 mixtures of metal hydrides.13 A crucial limitation of these results was that the lists of reactions we considered were generated “by hand”, and no guarantees could be made that the individual reactions would proceed as single-step processes. Several reactions we initially predicted to be interesting were later found to be relatively uninteresting because they proceed by multistep paths or via products that we had not considered.14 In this paper, we extend our earlier work in two significant ways. First, we screen >16 million distinct mixture compositions of crystalline solids containing Al, B, C, Ca, K, Li, Mg, N, Na, Sc, Si, Ti, V, and H, an analysis that exceeds all previous efforts by orders of magnitude. Performing the same search experimentally would require vast resources. Second, we use newly introduced methods14 to rigorously identify the preferred reaction pathway for each mixture. By combining this new methodology with what we believe to be the largest library of target compounds currently available, we present below the most comprehensive screening of metal hydride mixtures for H2 storage to date. © 2008 American Chemical Society

Letters We first describe the application of first-principles calculations to individual reactions of interest. The total energy of each crystalline compound we considered was calculated using plane wave density functional theory (DFT) calculations with the projector augmented wave method15 in conjunction with the PW91 functional16 after optimizing the compound’s crystal structure within its experimentally reported crystal structure17 (see the Supporting Information). When required, the compound’s vibrational density of states (VDOS) was computed within the harmonic approximation.18 The VDOS is required to compute the free energy of reaction at finite temperature, ∆G(T), because this quantity is influenced by zero point energies and T-dependent vibrational contributions. Comparing values of ∆G(T) computed in this way with experimental data for a range of metal hydrides and hydride mixtures showed that the uncertainty in the DFT results is (10 kJ/mol H2.8,13 This precision is sufficient to separate reactions with unacceptable decomposition temperatures from those with more favorable thermodynamics. Computing a solid’s VDOS is much more time consuming than a DFT total energy calculation. Fortunately, reaction thermodynamics for metal hydride decomposition can be estimated with acceptable accuracy using only total energy calculations.8,13 We refer to the reaction enthalpy computed in this way as ∆U0. This quantity systematically overestimates ∆G(T) at the temperatures relevant to H2 release, but the magnitude of this correction can be estimated reliably.8 ,13 As a result, we use a two-stage approach to our thermodynamic screening. First, the thermodynamics of the largest possible number of reactions is assessed using only total energy calculations using ∆U0. Second, (when feasible) ∆G(T) is calculated for promising individual reactions found in the first stage. Once a catalog of compounds is available, enormous numbers of reactions can be examined individually with the methods outlined above. Akbarzadeh et al. recently introduced an elegant means to rigorously determine the preferred reaction path among all possible combinations of compounds within a specific library of compounds.14 Briefly, at a specified temperature and H2 pressure, mixtures of compounds with a fixed total composition of non-H elements are considered. The mixture that minimizes the system’s grand potential at these conditions is the equilibrium state. By repeating this calculation at varying temperatures, reactions can be detected as changes in the equilibrium state. If an overall reaction proceeds via multiple steps, then these steps can be detected easily. Akbarzadeh et al. examined mixtures containing Li, Mg, N, and H. Below, we extend their analysis to a much broader collection of elements. All calculations we present below utilize this grand potential method, so all reactions we report are definitively single-step reactions (within the set of mixtures defined by our database of compounds). An example of the ability of our calculations to predict multistep reactions is the mixture Ca(AlH4)2 + 2MgH2. We performed grand potential minimization calculations using ∆U0 and considering all combinations of Al, Ca, H, Mg, Al2Ca, AlH3, Al12Mg17, CaMg2, Ca(AlH4)2, CaAlH5, CaH2, MgH2, and Mg(AlH4)2. This list includes all Al-Ca-H-Mg compounds with known crystal structures. Our calculations predict an overall reaction that occurs in four distinct steps. At low temperatures, Ca(AlH4)2 decomposes: Ca(AlH4)2 f CaAlH5 + Al + 1.5 H2. As T is increased further, CaAlH5 f CaH2 + Al + 1.5 H2. Only after this step does the MgH2 react via 12Al + 17MgH2 f Al12Mg17 + 17H2. Finally, our calculations predict that CaH2 decomposes via CaH2 + Al12Mg17 f 17Mg + 6Al2Ca + 6H2

J. Phys. Chem. C, Vol. 112, No. 14, 2008 5259 at temperatures in excess of 880 K. Recent experiments by Hanada et al. with ball-milled Ca(AlH4)2 + 2MgH2 showed the same sequence of reactions with the exception that the last step was not observed.19 It seems likely that the final step was not observed because the experiments only included temperatures below 720 K. Two crucial caveats are associated with our method. First, if the true reaction involved a compound whose crystal structure is not currently known, or if the reactants or products form amorphous or disordered materials, then our calculations could not predict this reaction. Second, our method can give no information about the kinetics of a particular reaction; we can only determine the overall reaction thermodynamics. Therefore, our method may predict a reaction that will not be experimentally observable because of kinetic limitations. We now describe our materials screening calculations. We performed DFT calculations for the 212 crystal structures containing Al, B, C, Ca, K, Li, Mg, N, Na, Sc, Si, Ti, V, and H listed in Table S.1 (see Supporting Information). This list includes all relevant compounds from Pearson, Wycoff, and ICSD.20-22 Importantly, it also includes the recently reported experimental crystal structure of Mg(BH4)2.23 The only gaseous compound we considered was H2; we did not include NH3 or gaseous hydrocarbons as potential products. Using the 13 non-H elements listed above, we examined all 715 element spaces of the form E1-E2-E3-E4-H, where Ei is an element from the list above. This approach includes analysis of all 286/98/13 spaces of the form E1-E2-E3-H/E1-E2-H/E1-H. In each element space, distinct compositions filling the entire space were defined using increments in the non-H mole fractions of 0.02. At each composition, the grand potential minimization method using DFT total energy values was applied as T ranged from 0 to 1000 K in increments of 5 K. In these calculations, the T dependence of the grand potential enters only via the chemical potential of H. For every reaction listed below, these calculations were repeated for verification using T increments of 0.5 K. In all, these calculations examined >16 million distinct mixtures. To use the calculations outlined above, selection criteria must be specified to separate the reactions with useful properties from other reactions. We did this by only retaining single-step reactions that release >6.0 wt % H at completion and for which ∆U0 lies in the 15-75 kJ/mol H2 range. This enthalpy range is inclusive in the sense that reactions excluded by this criterion will not have desirable reaction thermodynamics even after the uncertainties associated with the DFT calculations are considered. This approach identified 43 distinct reactions, many of which have not been studied previously. All of these reactions are listed in Table 1 and are also shown in Figure 1. Almost all of the reactions we identified involve the combination of either LiBH4 or Mg(BH4)2 with other materials. For most of the reactions involving LiBH4, the reduction in the reaction enthalpy relative to the direct decomposition of LiBH4 is modest. Mixtures of LiBH4 with TiN, TiH2, ScH2, or C, however, yield reactions with substantially lower reaction enthalpies. Some of these reactions involve products that are regarded as refractory, for example TiB2; these may imply the existence of severe kinetic limitations. All of the reactions we found that contain LiBH4 involve the combination of LiBH4 with only one other compound. For the reactions containing Mg(BH4)2, however, reactions involving both one and two additional compounds were found. Several reactant mixtures were found for which the DFT-calculated reaction enthalpy is 10-20 kJ/mol lower than that for the direct decomposition of Mg(BH4)2. Ten reactions that did not include either LiBH4 or

5260 J. Phys. Chem. C, Vol. 112, No. 14, 2008

Letters

TABLE 1: Complete List of Single-Step Reactions Identified from a 212 Compound Database With >6 wt % H2 Released at Completion and 15 e ∆U0 e 75 kJ/mol H2a reaction Mg(BH4)2 f MgB2 + 4H2 Si + 4Mg(BH4)2 f Mg2Si + 2MgB4 + 16H2 LiBH4 + C f LiBC + 2H2 6LiBH4 + CaH2 f 6LiH + CaB6 + 10H2 8LiBH4 + MgH2 + BN f 8LiH + MgB9N + 13H2 2LiBH4 + MgH2 f 2LiH + MgB2 + 4H2 3Si + MgSiN2 + 12Mg(BH4)2 f 4Mg2Si + 2MgB9N + 36H2 BN + 4Mg(BH4)2 f 3MgH2 + MgB9N + 13H2 NaH + 2Mg(BH4)2 f NaMgH3 + MgB4 + 7H2 CaH2 + 1.5Si + 3Mg(BH4)2 f CaB6 + 1.5Mg2Si + 13H2 2C + Mg(BH4)2 f MgB2C2 + 4H2 CaH2 + 3Mg(BH4)2 f 3MgH2 + CaB6 + 10H2 8LiBH4 + Mg2Si f 8LiH + Si + 2MgB4 + 12H2 2LiBH4 + ScH2 f 2LiH + ScB2 + 4H2 2LiBH4 + TiH2 f 2LiH + TiB2 + 4H2 2LiBH4 + NaMgH3 f 2LiH + NaH + MgB2 + 4H2 3NaH + BN + 4Mg(BH4)2 f 3NaMgH3 + MgB9N + 13H2 2LiBH4 + Mg(NH2)2 MgH2 + 2LiH + 2BN + 4H2 ScH2 + Mg(BH4)2 f MgH2 + ScB2 + 4H2 MgH2 f Mg + H2 5B + Mg(BH4)2 f MgB7 + 4H2 2MgH2 + Mg(NH2)2 f Mg3N2 + 4H2 CaH2 + 3NaH + 3Mg(BH4)2 f 3NaMgH3 + CaB6 + 10H2 6LiBH4 + 2ScN f 6LiH + 2ScB2 + 2BN + 9H2 3Si + 8BN + 5Mg(BH4)2 f 2MgB9N + 3MgSiN2 + 20H2 4LiBH4 + K2MgH4 f 4LiH + MgB4 + 2KH + 7H2 4LiH + 3Mg(NH2)2 + 2C f 2Li2CN2 + 2Mg3N2 + 8H2 6LiBH4 + 2TiN f 6LiH + 2TiB2 +2BN + 9H2 2LiNH2 + C f Li2CN2 + 2H2 Al + MgB9N + 2.5Mg(BH4)2 f AlN + 3.5MgB4 + 10H2 2LiBH4 + MgB2 f 2LiH + MgB4 + 3H2 MgB7 + 1.5Mg(BH4)2 f 2.5MgB4 + 6H2 12LiH + 3Mg(NH2)2 + 4BN f 4Li3BN2 + Mg3N2 + 12H2 K2MgH4 + 2Mg(BH4)2 f MgB4 + 2KMgH3 + 7H2 28LiH + 9Mg(NH2)2 + 4VN f 4Li7N4V + 3Mg3N2 + 32H2 2ScN + 3Mg(BH4)2 f 3MgH2 + 2ScB2 + 2BN + 9H2 NaH + ScH2 + Mg(BH4)2 f NaMgH3 + ScB2 + 4H2 4LiBH4 + 2KMgH3 f 4LiH + MgB4 + K2MgH4 + 7H2 2TiN + 3Mg(BH4)2 f 3MgH2 + 2TiB2 + 2BN + 9H2 2LiH + LiNH2 + BN f Li3BN2 + 2H2 2Li3Na(NH2)4 + 4C f 3Li2CN2 + Na2CN2 + 8H2 4LiH + 3LiNH2 + VN f Li7N4V + 5H2 10LiH + 5LiNH2 + N4Si3 f 3Li5N3Si + 10H2

∆U0 wt % (kJ/mol H2) H2 14.9 13.2 11.9 11.7 11.6 11.5 11.2

54.0 52.7 45.1 62.1 66.3 66.2 48.6

10.9 10.67 10.6 10.3 9.9 9.6 8.9 8.6 8.6 8.4

51.2 53.2 45.4 43.1 47.5 74.0 49.7 22.2 68.9 48.8

8.1 8.0 7.7 7.5 7.4 7.3

20.6 37.5 64.7 41.5 26.0 44.3

7.3 7.3 7.3 7.2 7.1 7.0 6.8 6.8 6.7 6.7

59.5 47.0 74.8 47.8 35.9 31.4 53.6 72.5 50.2 54.2

6.6 6.5

51.2 47.5

6.5 6.5 6.4 6.4 6.3 6.1 6.1 6.0

43.1 34.8 72.2 19.5 49.1 32.6 37.4 60.1

a All reactions that involve LiBH4 list ∆U0 calculated with the lowtemperature (ortho) crystal structure. Reactions are listed in order of decreasing H2 capacity.

Mg(BH4)2 met the selection criteria above. These include reactions involving Mg(NH2)2 from the mixtures 4MgH2 + Mg(NH2)2 and 4LiH + C + 3Mg(NH2)2. The outcomes of this screening exercise can be viewed in at least two ways. First, our results provide strong motivation to experimentally test the specific mixtures we have identified from our grand potential approach. At the same time, our calculations predict that no other combinations of the 212 compounds we have considered will allow single-step reactions to occur at the specified temperature and hydrogen pressure. After identifying the reactions described above using DFT total energy calculations, it is appropriate to undertake the more time-consuming VDOS calculations that allow ∆G(T) to be assessed. To illustrate this, we have performed these calculations for a subset of the reactions listed in Table 1. These calculations did not include a number of materials with relatively complex crystal structures. We did not perform VDOS calculations for

Figure 1. Reaction enthalpy as computed from 0 K DFT calculations, ∆U0, for all single-step reactions involving Mg(BH4)2 (filled squares), LiBH4 (open circles), and reactions that do not include LiBH4 or Mg(BH4)2 (triangles) satisfying the criteria defined in the text. The reactants that combine with LiBH4 or Mg(BH4)2 in these reactions are indicated in the figure. For other reactions, all reactants are shown. The error bars for each reaction span the predicted ∆G at 300 K assigned using the methods of refs 8 and 13.

Letters

Figure 2. Equilibrium pressures at 300 K for the 12 single-step reactions listed in Table S.3 computed as described in the text. For the 10 reactions involving LiBH4, only reactants other than LiBH4 are labeled in the figure. For the other two reactions (labeled by *), all reactants are listed in the figure.

Mg(BH4)2, for example, because of the complexity of the crystal structure, which contains 330 atoms per unit cell.23,24 Our calculations accounted for the ortho to hex phase transition in LiBH4 as described previously.8 After computing ∆G(T), we characterized each reaction by finding the temperature at which Peq ) 1 bar, Teq, and also the equilibrium pressure at T ) 300 K. Our predicted equilibrium pressures at 300 K are shown in Figure 2. The uncertainties shown in this figure result from associating an uncertainty of (10 kJ/mol H2 with our DFTcalculated free energies, consistent with the discussion above. This uncertainty corresponds to a large range of P and T because of the van’t Hoff relationship between P, T, and free energy.2,8,25 Table S.3 lists the reaction entropies at 300 K for the reactions shown in Figure 2. Except for the direct decomposition of MgH2, these entropies lie in a narrow range, consistent with previous discussions of this quantity.2,8 This observation justifies focusing attention on the reaction enthalpy in classifying reactions. Table S.3 also lists the range of transition temperatures obtained using ∆G(T) ( 10 kJ/mol H2. These results are consistent with the remark above that ∆G(T) is overestimated by ∆U0 and support the treatment in refs 8 and 13 that was used to give estimates for ∆G in Figure 1. The uncertainties that exist in the predicted transition temperautres stem from the inexact nature of the functionals available in current DFT calculations, and no obvious method exists to reduce this uncertainty that can be readily applied to a catalog of hundreds of crystal structures. It is clear from these uncertainties that precise experimental measurements of the equilibrium properties of the reactions we have identified will be crucial to make better informed decisions about their suitability for practical applications. The grand potential method we have applied here has several attractive features. As stated above, this method rigorously yields the equilibrium mixture composition with the set of all mixtures of compounds that are included in our database. Moreover, this minimization problem defining the method can be specified as a linear program,14 so its unique solution can be found extremely rapidly. The full set of screening calculations described above (>16 million distinct mixture compositions, each treated at 200 distinct temperatures) was performed in several days once the database of DFT total energies was available. This means that it will be possible to routinely repeat this complete screening process whenever the crystal structure of a new compound that is potentially relevant becomes available.

J. Phys. Chem. C, Vol. 112, No. 14, 2008 5261 In any effort to screen large numbers of materials for a complex application, it is vital to clearly understand the limitations of the screening methods and criteria. We have only retained single-step reactions as candidate reactions. Other criteria such as inclusion of multistep reactions that release an appropriate amount of H2 over a specified temperature range could also be considered. Our approach cannot predict the formation of materials if their crystal structure is not known and included in our database. This implies that efforts to synthesize and determine the structure of new complex metal hydrides remain crucial to the search for high-performance materials. Our approach does not include the possibility of doping metal hydrides with other species to modify their thermodynamic properties.26 For dopants to be useful in reactions involving chemical mixtures, multiple stability criteria must be satisfied. DFT calculations can be useful for seeking advantageous dopants once an interesting reaction involving stoichiometric compounds has been identified.27 Some potential H2 storage materials release small amounts of NH3 or other gases in addition to H2; this can create problems for contamination of the H2 that is released and with long-term materials stability.28 Our calculations do not make predictions about this situation because the only gaseous species they consider is H2. In summary, we have described a rigorous screening of metal hydride mixtures for high-capacity H2 storage based on the most complete database of crystal structures currently available. This effort identified 43 distinct single-step reactions that satisfied our selection criteria and, just as importantly, showed that no other combinations of the known reactants achieve these goals. The results represent an important step toward the enumeration of the largest possible set of materials whose thermodynamics meet the challenging targets required for vehicular H2 storage and can be used to great advantage in efforts to select materials that also achieve the multiple non-thermodynamic goals (e.g., reaction kinetics, cost) that must be met by any commercially viable material. Acknowledgment. This work was supported by the DOE grant no. DE-FC36-05G015066 and performed in conjunction with the DOE Metal Hydride Center of Excellence. Supporting Information Available: List of compounds considered in the calculations and their structural parameters and table of estimated transition temperatures for the reactions shown in Figure 2. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Zu¨ttel, A. Naturwissenschaften 2004, 91, 157. (2) Grochala, W.; Edwards, P. P. Chem. ReV. 2004, 104, 1283. (3) National Hydrogen Energy Roadmap; US Department of Energy; http://www.eere.energy.gov/hydrogenandfuelcells/pdfs/hydrogen_posture_plan.pdf, 2002; p 17. (4) Zhao, X.; Xioa, B.; Fletcher, A. J.; Thomas, K. M.; Bradshaw, D.; Rosseinsky, M. J. Science 2004, 306, 1012. (5) Lee, H.; Lee, J.-w.; Kim, D. Y.; Park, J.; Seo, Y.-T.; Zeng, H.; Moudrakovski, I. L.; Ratcliffe, C. I.; Ripmeester, J. A. Nature 2005, 434, 743. (6) Lee, H.; Choi, W. I.; Ihm, J. Phys. ReV. Lett. 2006, 97. (7) Vajo, J. J.; Skeith, S. L.; Meters, F. J. Phys. Chem. B 2005, 109, 3719. (8) Alapati, S. V.; Johnson, J. K.; Sholl, D. S. J. Phys. Chem. C 2007, 111, 1584. (9) Chen, P.; Xiong, Z.; Luo, J.; Lin, J.; Tan, K. L. Nature 2002, 420, 302. (10) Bogdanovı´c, B.; Schwickardi, M. J. Alloys Compd. 1997, 253254, 1. (11) Vajo, J. J.; Olson, G. L. Scr. Mater. 2007, 56, 8.

5262 J. Phys. Chem. C, Vol. 112, No. 14, 2008 (12) Sudik, A.; Yang, J.; Halliday, D.; Wolverton, C. J. Phys. Chem. C 2007, 111, 6568. (13) Alapati, S. V.; Johnson, J. K.; Sholl, D. S. Phys. Chem. Chem. Phys. 2007, 9, 1438. (14) Akbarzadeh, A.; Ozolinsˇ, V.; Wolverton, C. AdV. Mater. 2007, 19, 3233. (15) Kresse, G.; Joubert, D. Phys. ReV. B 1999, 59, 1758. (16) 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. (17) Alapati, S. V.; Johnson, J. K.; Sholl, D. S. J. Phys. Chem. B 2006, 110, 8769. (18) Ackland, G. J. J. Phys.: Condens. Matter 2002, 14, 2975. (19) Hanada, N.; Lohstroh, W.; Fichtner, M. J. Phys. Chem. C 2008, 112, 131. (20) Villars, P. Pearson’s Handbook: Crystallographic Data for Intermetallic Phases, Desk ed.; ASM International: Materials Park, OH, 1997.

Letters (21) Wycoff, R. W. G. The Structure of Crystals; The Chemical Catalog Company Inc.: New York, 1931. (22) The Inorganic Crystal Structure Database (ICSD), http://www.fizinformationsdienste.de/en/DB/icsd/. (23) Cerny, R.; Filinchuk, Y.; Hagemann, H.; Yvon, K. Angew. Chem., Int. Ed. 2007, 46, 5765. (24) Dai, B.; Sholl, D. S.; Johnson, J. K. J. Phys. Chem. C, in press, 2008. (25) Vajo, J. J.; Mertens, F.; Ahn, C. C.; Robert C.; Bowman, J.; Fultz, B. J. Phys. Chem B 2004, 108, 13977. (26) Streukens, G.; Bogdanovic, B.; Felderhoff, M.; Schuth, F. Phys. Chem. Chem. Phys. 2006, 8, 2889. (27) Alapati, S. V.; Johnson, J. K.; Sholl, D. S. Phys. ReV. B 2007, 76, 104108. (28) Chen, P.; Xiong, Z.; Yang, L.; Wu, G.; Luo, W. J. Phys. Chem. B 2006, 110, 14221.