In our search for new and improved hydrogen storage materials, we are guided by advanced theoretical predictions of the kinetics and thermodynamics of potential hydrogen releasing reactions. Obtaining accurate enthalpy estimates of potential new hydride decomposition reactions is a necessary means of identifying which materials systems we should attempt to synthesize. Such predictions require input of the crystal structures of the chemical reactants. Sometimes, these structures are already known experimentally. However, for those cases where the structures are unknown, the ability to predict the crystal structure is required. The predicted structures can then be used in first-principles calculations of thermodynamic properties of hydrogen absorption and desorption.

As part of our MHCoE work we have developed a Monte Carlo (MC) global optimization algorithm to perform crystal structure prediction of complex anionic hydrides. These compounds consist of molecular anionic units such as (AlH₄^-, (AlH₆^3-), (BH₄^-), (NH^2-), and (NH₂^-), charge balanced by a cation matrix consisting of the alkali or alkaline earth metals. The predicted crystal structures are used in first-principles calculations of reaction enthalpies.

Our global optimization algorithm utilizes an energy functional consisting of simple electrostatics and distance scaling (DSM) to smooth the potential energy surface (PES). We attempt to find the electrostatic ground state of a collection of anions and cations using this DSM-MC approach. First-principles Density Functional Theory (DFT) methods are used to calculate quantum mechanical T=0 K energies from the DSM-MC algorithm output. Structural relaxation of the DSM-MC generated structures is performed using standard DFT techniques. These first-principles energies are used to obtain enthalpy estimates at 0 K. Lattice vibration (phonon) contributions to the enthalpy can be calculated for finite temperature effects where necessary.

Figure 1: First principles energies for two formula units of Ca(BH₄)₂ calculated from single run MC (black circles) and DSM-MC (red circles) methods. The initial energy (abscissa) is the calculated DFT energy before structural relaxation of atom positions and lattice parameters. The DSM-MC generated structures show a clear and dramatic improvement over single-run MC minimization.

The algorithm correctly produces the known ground state structures of NaAlH₄, Mg(AlH₄)₂, and K₂LiAlH₆, for example, illustrating the success of the code. In addition, the DSM-MC method is competitive with database searching methods, and has produced unsuspected high symmetry structure types not found in large crystal structure databases such as the Inorganic Crystal Structure Database (ICSD). A publication detailing our results is in preparation. These results give us confidence that the DSM-MC approach is producing crystal structures that are likely very close to ground state energies, and therefore that the enthalpy estimates derived from the subsequent calculations (ignoring phonon contributions to the free energy) may be accurate to within about 10 kJ/mol. We continue our search for bi-alkali borohydride systems, including but not limited to, Li-Mg, Li-Ca, Li-Na, Li-K, Na-Mg, Na-Ca, Na-K, K-Mg, K-Ca.

Contact:

Eric Majzoub, Sandia National Laboratories, currently on leave at University of Missouri, St. Louis
Phone: (314) 516-5779
Email: majzoube@umsl.edu