Research and practical problem-solving unite for David Keffer and Brian Edwards, the co-leaders of UT's Computational Materials Research group, whether they're developing a machine lubricant that won't lose its effectiveness in Arctic climates, clarifying molecular processes inside a hydrogen fuel cell, or figuring out how flow and resistance affect molded plastic.
Both scientists use high-end computer simulation to study dynamic molecular relationships within a material, searching for molecular arrangements that will operate within the constraints imposed by the desired function and the environment in which it will be used. "We're engineers and we use computers to solve practical problems," says Keffer.
In his office, surrounded by his young daughter's colorful artwork, Keffer explains the timesaving advantages of their approach. "We can do rapid screening of different compounds to determine if it's worth the time and effort to move on to experimental synthesis and characterization," he says. "Sometimes we see quickly that this is a dead end; other times, we see real promise for developing something useful." Another advantage of their computer-based approach is the chance to observe how interactions among the atoms and molecules contribute to viscosity and other properties of a material. "Using computerized tools, we can number the molecules and freeze the simulated experiment at any instant," Keffer says.
To ensure that the simulations correspond to reality, Keffer and Edwards collaborate with experimentalists, among them UT chemistry professor Jaimie Adcock, who's synthesizing compounds for the Arctic lubricant simulations, and colleague Bill Steele, who's testing the compounds in the UT Thermophysical Measurement Laboratory.
Edwards insists that to be useful, simulations that mimic real experiments not only must reproduce experimental conditions, they must also calculate the same properties that will be measured in an actual experiment. That's a tall order, he acknowledges, due to limitations inherent in the simulation process. "Simulation is done with mathematics, not matter, and mathematics can introduce artifacts that destroy what you're trying to describe," Edwards says. "Also, the computational time required to do a simulation forces us to use much smaller numbers of particles [in the thousands] than would exist in actual experimental samples [billions upon billions]."
In devising a typical simulation, Edwards's key task is to develop algorithms that conform to the laws of physics, a skill he has been perfecting for 15 years. That's when Edwards and other researchers began working out algorithms to take into account—or conserve—the total energy within a group of interacting molecules, whether the energy comes from temperature, particle velocity, or the position of particles in a material. "Many times we can drop the simulation data right on top of the experimental data," he says, "with the statistical error in the simulation the same as that in the experimental measurement."
Despite the exacting precision required for his work, Edwards enjoys the potential for discovery that surrounds his efforts. "The longer the problem has been around, the more exciting it is to discover the solution," he says.
Meanwhile Keffer acknowledges the creativity that is central to his work and how that creativity often is spurred by interacting with his colleagues. "All it takes is an off-hand comment by somebody to prompt you to think about a problem in a different way," he says. Keffer commutes to work on a bicycle, and as he pedals, the wheels in his head often begin to turn, as well: "I'll think about a conversation I had with a graduate student or a post-doc," he says, "and I suddenly realize That's exactly right! Why didn't I think of it myself? When he arrives at his lab with the eureka moment still fresh in his mind, Keffer fires up his computers and begins to use numbers to build something that will—if all goes well—evolve into a practical, and very real, material.
- - -