It takes a big computer to model very small things. And, like its namesake state, New York Blue is big. Made up of 36,864 processors, the massively parallel IBM Blue Gene/L is housed at DOE’s Brookhaven National Laboratory (BNL) on New York’s Long Island, where, among other things, it’s used to model quantum dots, or nanoparticles, just a few atoms in size.
Overseeing New York Blue’s care and feeding is Brookhaven’s Computational Science Center, headed by James W. Davenport. Davenport is spearheading an effort to model nano-sized slivers of metallic material, such as gold and palladium, on the supercomputer.
Modeling nano-sized particles is necessary because conventional diagnostic tools have not yet been able to accurately determine the makeup of such tiny bits, which differ from larger chunks of the same materials. When large pieces of gold and palladium are X-rayed their crystal structure becomes apparent. “The atoms are arranged in a certain periodic array that is well defined. But small particles are not the same. It is not known exactly what the atomic structures of most of these particles are,” Davenport says.
Learning the physical properties of these nanocrystals is the first step in determining how they may be used as catalysts — materials which manipulate chemicals — and in turn create new applications in DOE-related fields such as advanced energy technology.
Hydrogen storage turns golden
Gold, for example, is a noble inert metal that normally does not react with other chemicals. Yet when gold particles made up of fewer than about 1,000 atoms are placed in the presence of certain chemicals, catalytic reactions take place.
A particular catalytic interaction involving gold nanoparticles may prove useful for producing hydrogen, most of which is currently made from natural gas. Hydrogen separated from natural gas often also contains carbon monoxide, which poisons the fuel cells used to produce electricity. Gold nanocrystals may be used as a catalyst to scrub carbon monoxide from the hydrogen stream. This could greatly expand clean electricity production, since water is the only byproduct of fuel cell operation.
There also is evidence that nanoparticles might be a good hydrogen storage medium. Hydrogen storage currently requires considerable space, presenting a major obstacle to its use as a significant alternative energy source. Storing hydrogen in a large tank, for example, makes it less practical as an automotive fuel. Cutting the space requirement by getting hydrogen to adhere to tiny particles of metal would be a major step toward creation of a hydrogen-based economy.
All of this makes it important “to understand how the chemical bond of hydrogen to these nanoparticles differs from the chemical bond of hydrogen to a large chunk of metal,” Davenport says.
To accomplish these goals first requires understanding how the properties of the very small differ from the big. And that is where the modeling on New York Blue comes in.
Searching for low energy
“We are studying the shapes of the different particles as a function of their size by calculating the energies of these different structures and seeing which one is the lowest,” Davenport says.
The energy in this case is potential energy — the force drawing the atoms together — as opposed to kinetic energy, which involves actual movement of the atoms. “If I allow two atoms to come close together the bond energy increases and that’s what makes chemical bonds — that’s what holds our world together,” Davenport says. “We are trying to calculate that chemical bond energy.”
Researchers look for structures with the lowest energy because that’s how nature tends to arrange things. To conduct this search, Davenport’s team computationally arranges atoms into different shapes and models their potential energy.
“If two atoms are far apart then the energy is two times the sum of the isolated atoms,” Davenport explains. “If I allow those two atoms to come close together they will form a chemical bond. It is the chemical bond energy that we are really looking for.”
As part of their search, the investigators use quantum mechanics to simulate the arrangement of atoms in different configurations and then attempt to determine which has the lowest energy.
“If I have three atoms I can put them into a triangle. If I have four I can put them into a tetrahedron, or if I have eight I can put them into a cube,” Davenport says. “I can then vary the cube edge to see how the energy varies. If the cube edge is very small, usually the energy will be high, because these atoms are too close together and repel each other. If I take them far apart they also have a high energy because they like to bind together. So someplace in between there is an optimum and we search for those optima.”
Scaling up
Researchers at BNL’s Computational Science Center have modeled nanoparticles for several years and succeeded in calculating the crystalline structures of small clusters of up to 150 atoms of gold and palladium. Their results agree with some empirical evidence.
Up to now, the group has verified its results with cutting-edge optical experiments that are slowly becoming available. “There are molecular beam experiments in which these particles can sometimes be created and can be stuck on surfaces and examined with electron microscopes, but it is difficult,” Davenport adds. The researchers also have looked at hydrogen as though it was stuck on a palladium nanostructure, but do not yet have experimental confirmation of their results. They’re now working to scale up their models to examine the structures of particles consisting of hundreds to thousands of atoms.
The researchers chose to model metals first because of previously existing evidence that they would make good storage media and because of past experience at BNL, where other groups have modeled metals for other potential applications.
Davenport’s team has used a number of DOE-developed computer codes in its work, including NWChem, which was developed at Pacific Northwest National Laboratory; GAMESS, a product of Ames Laboratory; and a home-grown code dubbed LASTO.
In addition to the nanoscience work, other researchers at BNL’s Computational Science Center are working on magnetohydrodynamics, with a particular focus on ITER, a fusion research project undertaken by an international consortium including the United States. The ITER reactor, being built in France, will be used to create and study conditions required to produce electricity through nuclear fusion, rather than fission, the process behind present nuclear power plants. Nuclear fusion, the process that takes place in the Sun, holds the potential for solving the world’s energy problems because it would be a virtually limitless energy source independent from finite fossil fuels.
Brookhaven researchers also are involved in modeling the intense collisions produced in Brookhaven’s Relativistic Heavy Ion Collider (RHIC). Activated in 2000, the RHIC is used to study what the universe may have looked like in the first few moments after its creation. It does so by crashing beams of gold ions head-on in a subatomic collision. Davenport’s department is helping increase the intensity of the beam in the collider by simulating the motion of particles in the accelerator.
Some of this computational science work is done on other computers at the Computational Science Center, including the 500-processor Galaxy, a conventional Beowulf cluster. But it is New York Blue that has become the shining star of a very small universe.