Analyzing bottlenecks in chemical reactions
His goal is to determine with greater accuracy the nature of the bottlenecks, ultimately providing scientists with a far greater comprehension of important reactions, allowing the researchers to better control them. Fields that are expected to benefit from this research range from atmospheric chemistry to automotive engine design.
When it comes to understanding the spectrum of chemical reactions, King says highly accurate data are needed for just a few key reactions.
"Just as there are multiple routes that will get you from UB to Kleinhans Music Hall near downtown Buffalo, there are multiple ways to get from reactants to products," explains King. "Molecules may encounter minor traffic jams while approaching and leaving the bottleneck, but these are of little importance compared with the time it takes to get through the one particular elementary reaction that has a high-energy barrier: the bottleneck itself."
Identifying accurate rates for these key reactions, which involve the breaking of a chemical bond and the formation of a new one, is critical. During a chemical reaction, King notes that molecules go through a transition state.
"To predict the speed of the reaction, one needs to know the shape and energy of the molecule in the transition state," says King. That's easier said than done, however.
Even using state-of-the-art experimental techniques, it is usually impossible to observe molecules going through the transition state since they exist in that state for only a fraction of an instant. However, by executing extremely complex calculations based on the theory of quantum mechanics, King says, super-computers allow computational chemists to predict the energy and structure of these important, but fleeting, intermediate molecules or molecular fragments.
"As computational chemists, we treat molecules like mechanical systems that consist of particles, nuclei and electrons, in order to examine the mathematical relationships among them," explains King. "If you can ‘solve' those relationships, then you could, in principle, answer almost any question."
He notes, however, that scientists never really solve these equations. "We always make mathematical approximations," he explains. "These approximations have gotten awfully good over the past few years, but we'd like to make them even better."
With funding from the National Science Foundation, King and his colleagues at UB are developing a method that they hope will make those approximations from 10 to 100 times more accurate, a goal they hope to attain more quickly thanks to the power of the Dell cluster.
As a first test of the new method, the UB team is studying a simple molecule composed of two carbon atoms, which has the distinction of being a stable molecule, so it can be studied in the laboratory; it also has many similarities to the fleeting intermediate species that occur in the very short-lived transition state that chemists long to study.
"It has a lot in common with a typical transition state, so we picked it out as a nice test case," says King. "Yet even with such a simple molecule, this problem is too big to run on a single processor."
King's team has found that it takes several days to complete one run using a dozen processors on one of the UB Center for Computational Research's parallel computers. "On the new Dell cluster, instead of taking several days, it takes just a few hours," King observes.
In order to solve all of the equations related to the dissociation of the two carbon atoms, the UB team developed a mathematical expression called a "wave function" that contains a whopping 140 million terms.
Their findings generated on the Dell cluster will be compared to the findings obtained by laboratory spectroscopic analyses, which measure the diatomic carbon molecule. This brings the UB team closer to developing a robust method that would allow scientists to quickly figure out the few terms that are critical to determining the reaction rate and the mechanisms for a particular reaction. As for how long it will be before the team develops such a method, King estimates it could take just a few years.
"Between the calculations we're making using the supercomputers and the observations that have been made in the lab," he says, "we're getting fantastically good agreement."
return to High-speed science main page