UT Chemist Graeme Henkelman uses Ranger to test a system of nanotech prediction: "Lately, there’s been a lot of excitement about nanotechnology,” Graeme Henkelman, professor of chemistry at The University of Texas at Austin, said, with characteristic earnestness. “People have realized that as you make things small, particularly on the nanoscale, there are some properties that come out that are completely different than the bulk materials.” A transistor that worked well at ten nanometers will display odd quantum effects or heat up immensely at two nanometers. And a classically inert material like gold, when reduced, will suddenly become highly reactive. “You see very different chemical properties, and then you think, what are the other things that you could make small? If we could understand the nature of these materials, perhaps we could design them.” Since the first discoveries in nanoscale materials science a decade ago, furious experimentation has led to hundreds of amazing discoveries, from flash memory cards to dirt-repelling pants. But the knowledge has been piecemeal. Henkelman wondered if there might be a way to predict in advance — using the simulating power of emerging petascale supercomputers — how new materials, constructed one atom at a time, might perform. With that goal in mind, he has pursued the development of a method to simulate nanoscale systems two to three nanometers across (or approximately 200-300 atoms in size) to predict how these nanoscale materials will behave.
A model of a promising Paladium-shell/Cobalt-core nanoparticle for the catalytic reduction of oxygen in fuel cells.
Though nanotechnology has potential in numerous fields, from biology and medicine to electronics and optics, the applications that really interest Henkelman are in energy production, where he believes new technologies are required to replace the world’s rapidly diminishing supply of fossil fuels. “We want to understand at a fundamental level how you go from structure to function, and then build upon that,” Henkelman explained. “If we can understand the basics, then we can design things in an intelligent way.” Energy — whether from gasoline or electricity — comes largely from the breaking of bonds, but these chemical reactions don’t happen spontaneously. They need the right environment, the right conditions, and the right catalyst to change the chemical composition of stable molecules and generate energy. This is particularly true with hydrogen and oxygen, where the inert gases need a catalyst to expedite their interactions and create energy. Platinum, the material used for present-day fuel cells, is incredibly expensive, in limited supply, and not particularly efficient. But what nanoscale material would work better? And how can scientists learn the answer without testing every material, and combination of materials — a lengthy and expensive prospect? Building on past research, Henkelman is applying his system of computational simulation to what he believes are the most promising catalysts for hydrogen fuel reactions: core-shell bi-metallic nanoparticles, where a core of one metal is covered with a one-atom-thick shell of another metal. In these dual metals, the transfer of electrons between the two materials facilitates the reaction.
An oxygen molecule (red) reacts with a calcium atom (grey) on a magnesium oxide surface. Independent saddle point searches are used to find the important rare-events which govern the dynamics here and in many other chemical and materials systems.
“We pick two metals, one for the outside and one for the inside, and we use Ranger to calculate the structure,” Henkelman explained. “We do this by solving equations for where all the electrons are, using Density Functional Theory. Then we take small molecules like hydrogen and oxygen and calculate where they react on the surface of the nanoparticles, and what sort of energy barriers you have to go over for these reactions to take place.” The lower the energy barrier, the better the catalyst. Simulating bimetallic nanoparticles with a palladium shell (a less expensive relative of platinum), Henkelman and his graduate student, Wenjie Tang, are testing the catalytic potential of a number of different cores. Some, like cobalt and molybdenum, he believes will work well; others, like gold, shouldn’t work at all. By modeling a range of materials, and then comparing his results to the actual particles, he can determine whether his general methodology will succeed. “We’re sticking our neck out and saying, ‘This is our prediction. We think this material is going to be better than that material.’ Then, we pass it to the experimentalists and ask them to try this particular combination and see what they get.” Henkelman said. “A necessary step for this science to move forward is to make it predictive.” UT Chemist Dick Crooks sees great potential in Henkelman’s approach. Crooks developed a unique method for constructing real versions of the ultra small materials Henkelman simulates on Ranger. “Right now, new catalysts are discovered using chemical intuition or rapid screening of large libraries,” Crooks said. “It would be much better if we could calculate the structure of a catalyst from first principles.” Crooks is presently testing whether Henkelman’s predictions for the dual-metal systems are correct. Simulating the behavior of atoms is no easy matter — 99.999% of the time, atoms just vibrate. So a big focus of Henkelman’s research involves finding the rare but important reaction mechanism, when an oxygen atom dissociates or combines with hydrogen. Based on the energy landscape and the probability of a certain reaction occurring, he can calculate the rate of a specific reaction. However, so much is unknown about these nanoscale systems that calculating only the expected reaction isn’t adequate. “We simulate these systems over and over again on parallel processors to find all the potential outcomes. And these can be things we expect, like the formation of water, or things we don’t expect, like the particle changing shape,” Henkelman explained. “This is how we’re taking the science forward. Instead of just assuming we know what can happen, we’re using computer simulations to tell us what the dynamics are.” It is only with the advent of petascale computers like Ranger that it is even possible to simulate realistic-sized particles. “On Lonestar [TACC’s previous most powerful system], we could handle fifty to sixty atoms — under one nanometer. But on Ranger we simulate up to 200 atoms, which is what the experimentalists work with,” he said. To do so, Henkelman will use over 700,000 computing hours on Ranger in the coming year alone. If his predictions are borne out by experiment, Henkelman can move on to the next challenge: testing new combinations of promising materials. Over time, he hopes his computational method will guide experimentation and design in a rational direction, leading to powerful, efficient, ultra-small fuel cells, photovoltaics, and as-yet-unimagined energy solutions. “There are any number of materials out there, and to the extent that we see some success in our theory, this project could be very open-ended,” Henkelman said. “This is the start of a new field. It could take a lifetime, a whole career.” ***************************************** For more information, visit the
Henkelman Research Group website. To see an archive of nanoscale discoveries, visit
The Project on Emerging Nanotechnologies. Henkelman's recent paper in the Journal of Chemical Physics, "Adaptive kinetic Monte Carlo for first-principles accelerated dynamics," is available
online. Or to dig deeper into "Charge redistribution in core/shell nanoparticles to promote oxygen reduction," read the pre-print of his
most recent paper. Aaron Dubrow
Texas Advanced Computing Center
Science and Technology Writer
