Session W32: Focus Session: Frontiers in Computational Thermodynamics of Materials I

Modeling the interactions of adsorbates with each other and with metal surfaces

The interactions of molecules with metallic surfaces are fundamental to the ability of metals to catalyze reactions. One often thinks of a metal like platinum as the catalyst, but under reaction conditions the reactivity of the metal surfaces is modified by the molecules that adsorb on them. We have used quantum chemical calculations in conjunction with cluster expansions to probe the adsorption behavior of atomic adsorbates such as C, N, O, and S on late transition metal surfaces such as Rh, Ir, Pd, Pt, Cu, Ag, and Au(111). There are remarkable similarities in the adsorption behavior of these adsorbates that can be interpreted in terms of a simple adsorbate-induced surface electronic structure modification mechanism that is common to all the adsorbates and surfaces. The variations between the adsorbates and metals are readily explained in terms of the size of the metal and adsorbate orbitals and the geometry dependent overlap of these orbitals. We have constructed a new Solid State Table of these orbital radii from the quantum chemical calculations that can be used in conjunction with a simple model to rapidly estimate the electronic structure of metal and alloy surfaces with adsorbates on them.   

An Automatic Symmetry-Leveraging Approach for Solving Incomplete Many-Atom Crystal Structures

We present a new first principles-based method, called a symmetry-leveraging genetic algorithm (SLGA), for fully and automatically solving large crystal structures when experimental diffraction studies do not identify all internal atomic positions. Such incomplete structural refinements may occur when crystals contain light atoms or when the characterization is performed under extreme conditions such as high pressure. We apply our method to solve the crystal structure of the promising hydrogen storage candidate magnesium imide (MgNH), which has remained a mystery for over 40 years. We also confirm \emph{via} a fully automated procedure a recent specialized ``by hand" prediction for the high-pressure phase of ammonia borane, NH$_3$BH$_3$. The MgNH prediction, which involves 36 atoms and a notoriously complex configuration space, to the best of our knowledge represents the largest-ever crystal structure solution derived from first-principles calculations without making simplifying assumptions about atom connectivity. The 32-atom NH$_3$BH$_3$ prediction is nearly as demanding. Our approach, which takes full advantage of existing experimental information to solve for structural unknowns, has great potential for completing thousands of partially determined crystal structures.   

Finite Temperature Lattice Vibrations and the Magnetic Structure of Fe and Ni

Modern \textit{ab initio} theories of the magnetic phase transition (Curie Temperature, T$_{C})$ of Fe and Ni based on the Disordered Local Moment (DLM) type models generally rely on (constrained) density functional theory calculations performed at 0K and assume that the atoms occupy their equilibrium lattice sites. Here we point out that finite temperature lattice vibrations can result in large fluctuations in the local moments associated with individual site beyond those already accounted for in these approaches. These conclusions are based on large cell ($\sim $10$^{4}$ -- atoms) \textit{ab initio} calculations of the magnetic state of Fe and Ni based on the O[N] Locally Self-consistent Multiple Scattering (LSMS) method. Atom positions are obtained from freezes of individual time steps of molecular dynamics simulations based on classical interaction potentials. Calculations are performed for a range of temperatures up and beyond T$_{C}$ that illustrate the extent of the moment fluctuations. We discuss the consequences of these findings for the adequacy of existing theories T$_{C}$.   

Combined Experimental and Theoretical Studies of Core-Shell Nanostructures in Al-Sc-Li Alloys

We have used two aging treatments of Al-Li-Sc alloys to create highly monodisperse, coherent L$1_2$ structure core-shell precipitates. We perform detailed analyses of the compositional distributions in the precipitate structures with electron microscopy and atom probe tomography. By combining this information with first principles calculations and Monte Carlo simulations based on the cluster expansion formalism, we compute bulk and interfacial thermodynamic properties relevant to precipitate formation. We specifically focus on understanding how the presence of Li modifies the nucleation rate relative to that of pure Al3Sc precipitates.   

New structures in Pd-rich ordered alloys

An intriguing intermetallic structure with 8:1 stoichiometry was discovered in the 1950s in the Pt-Ti system. Since then, a handful of other Pt/Pd/Ni binary systems have been observed to exhibit this curious structure (e.g., Pt$_8$Zr, Pd$_8$Mo, Ni$_8$Nb, etc). This ordered structure can significantly increase the hardness of an alloy by forming precipitates. Recent calculations and experiments suggest that the 8:1 structure may form in about 20 previously unsuspected Pt/Pd binary systems. Using first-principles calculations, cluster expansion, Monte Carlo modeling, we have explored possible precipitate hardening (via the 8:1 structure) in Pd-Nb, Pd-Mg and Pd-Cu.   

Pseudo-Random Number Generation for Brownian Dynamics and Dissipative Particle Dynamics Simulations on GPU Devices

Brownian Dynamics (BD) and Dissipative Particle Dynamics (DPD) are implicit solvent methods commonly used in models of soft matter and biomolecular systems. The interaction of the numerous solvent particles with larger particles is coarse-grained as a Langevin thermostat is applied to individual particles or to particle pairs. The Langevin thermostat requires a pseudo-random number generator (PRNG) to generate the stochastic force applied to each particle or pair of neighboring particles during each time step. In a GPU parallel computing environment, small batches of random numbers must be generated over thousands of threads and millions of kernel calls. We introduce a PRNG scheme, in which a micro-stream of pseudorandom numbers is generated in each thread and kernel call. These high quality, statistically robust micro-streams are more computationally efficient than other PRNG schemes in memory-bound kernels, and uniquely enable the DPD simulation method . This scheme has been implemented in HOOMD-blue, a GPU-accelerated open-source general purpose molecular dynamics simulation package. By enabling BD and DPD to be performed in HOOMD-blue, a broad range of mesoscale coarse-grained simulations can now be accelerated in a massively parallel architecture.   

Efficient ab initio molecular dynamics using exact reweighting

Density-functional theory (DFT) based {\it ab initio\/} molecular dynamics (AIMD) is a promising method for calculating high-temperature thermodynamic properties of solids and liquids. Nevertheless, computational expense associated with AIMD simulations has prevented general adoption of these methods. We show that substantial savings of computational effort can be realized by using less expensive (and less accurate) Hamiltonians to generate long MD trajectories and by recalculating statistically independent snapshots with high-accuracy DFT methods. A formally exact reweighting formula, based on the Jarzynski switching approach, is used to obtain thermal averages and thermodynamic properties in the high-accuracy DFT ensemble. If under-converged AIMD simulations with low energy cutoffs and coarse k-point meshes are used to generate trajectories, this approach can lead to savings of CPU time of a factor of 10 to 100, depending on the relevant correlations times. We also present extensions of the reweighting method to calculate impurity free energies and free energy barriers for interstitial diffusion. Robust methods for estimation of statistical errors based on random subsampling and variance extrapolation are discussed.   

Ab Initio Simulations of Hydrogen in Crystalline and Amorphous Metal Membranes

Solid metallic membranes are used to separate hydrogen from other gases for clean energy applications. In order to create cheaper, more effective membranes for hydrogen separation, it is desirable to model hydrogen transport through the membrane. Amorphous metal membranes in particular have potential for this type of application due to low expense and high theoretical hydrogen capacity. We computationally model hydrogen absorption and transport through materials in order to find materials that can be used to construct effective membranes for hydrogen capture. In this talk, we will obtain hydrogen binding sites and diffusion barriers in order to model the hydrogen diffusion through various nickel-based amorphous alloys and compare them to associated crystalline structures as well as elemental palladium, which is favored for this application despite its high expense. Ab initio methods (specifically the Vienna Ab Initio Simulation Package, VASP) are used to develop the hydrogen binding energy spectrum, from which thermodynamic models can be constructed. Kinetic Monte Carlo methods are used to model the hydrogen transport through the bulk, from which we can obtain the permeability.   

The enigmatic Ag-Pt phase diagram and yet another derivative structure algorithm

The Ag-Pt phase diagram as published in the most recent phase diagram compilations (Massalski, Pauling File) is entirely speculative below 1000$^{\circ}$ C. The phase diagrams and our calculations both suggest a stable L1$_1$ phase at 50 at.-\% Pt. However, an experimental study published after the compilations supports a significantly different phase diagram. In this new phase diagram, the only stable phases at low temperatures are the elemental fcc Ag and Pt phases and one ordered phase at the unusual concentration of 53 $\pm 0.5$ at.-\% Pt. The experimental study shows that the homogeneity range for the ordered phase is narrow (almost like a line compound), and its X-ray data suggests that the unit cell of this phase contains 32 atoms with a stoichiometry of 15:17. We developed a new derivative structure enumeration algorithm specifically designed for large unit cells with known concentrations. This is necessary because our old algorithm enumerated all concentrations and was therefore limited to smaller unit cells. We have explored, via first-principles, the structural details of this enigmatic phase in the Ag-Pt phase diagram. I will discuss our first-principles results for Ag-Pt, and I will discuss the how the new algorithm is useful for large unit cells when partial structural information is known.