Modeling Palladium-Hydride on the T3D

To describe the interactions among the atoms in palladium-hydride, the WSRC team uses a method, the embedded atom method (EAM), originally developed at Sandia National Laboratory. The WSRC group, however, made an important revision. Unlike prior approaches, they allow the volume of palladium atoms to expand as the number of hydrogen atoms increases, thereby tracking a unique physical reality that occurs in the palladium-hydride system.

"Volume expansitivity is critical," says Wolf. "What we needed was a method to fix the chemical potential of hydrogen atoms, the number of palladium atoms, temperature and pressure. The talents of Professor Ray were invaluable." The final result is a simulation that allows hydrogen atoms to be created and destroyed in an expandable metal lattice.

(143 KB)

Onset of the beta phase in palladium hydride at 300 degrees Kelvin. This phase change occurs as the concentration of hydrogen atoms (yellow) in the palladium (purple) increases. At early stages (the alpha phase), hydrogen atoms randomly populate small interstices in the lattice structure. At a critical point, the lattice expands, allowing hydrogen to cluster at higher density, as visualized here. This image shows the lattice from the (001) direction.

In recent work, the group applied their method on the T3D at Pittsburgh (and at the Arctic Research Supercomputing Center). Vance Shaffer of Cray Research helped convert the code to run on the T3D. The results have been gratifying. "The T3D is a very stable machine," says Mansour, "and the performance has been wonderful. We've been able to simulate systems roughly 20 times larger than we've been working with till now." Their simulations of palladium with dislocations and grain boundaries show increases in the hydrogen concentration in the metal that correspond to expansion of the palladium atoms as hydrogen pressure goes up. 

These results, and the potential for even faster turnaround with better optimized code (which they're working on), have encouraged the WSRC group. They can now model systems of a size that approach reality, a factor especially important, says Wolf, to understanding grain boundaries and dislocations. "The systems we're running now are 3,000 to 5,000 atoms, and this range will allow us to investigate the role of defects in the materials."

A related goal that now appears in sight is to simulate alloys that could prevent or retard fracturing. One possibility, says Wolf, would be to alloy palladium with nickel at grain boundaries: "Hydrogen, as far as we know, doesn't like nickel, so it wouldn't go near the grain boundary. You can map out scenarios like that in your head, but you'd like to be able to prove them with simulations. Much of the experimental strategies in metal hydrides research are trial and error. A pinch of this, a tad of that. It's like the lottery -- you could get lucky. What we really need is better knowledge, and that's what simulations can give us."

go back to the main page