INDUSTRY

A new MIT-developed computational method for molecular modeling aims to help scientists reduce soot in hydrocarbons, understand what causes ozone degradation, and discover other chemical molecular behaviors that could potentially improve the atmosphere. Using Interactive Supercomputing Inc.'s (ISC) Star-P software, researchers at MIT's Department of Chemistry discovered a new method for calculating thermochemical properties derived from first principles, fundamental laws of chemistry that are not derived from any other source. The method is significant because, until now, scientists have struggled to calculate the energy surfaces, equilibrium and reaction rates of molecules at varying temperatures because they lacked accurate models. Instead, research has relied on conventional approximations of molecular behavior. A technical paper about the method was published last month in the Journal of Computational Chemistry. Star-P was instrumental in the research because the massive data and mathematical models required parallel processing capabilities, yet scientists also needed to be able to quickly interact with and manipulate the resulting data. "We needed supercomputing capabilities with very large memory arrays, but programming the code in C, FORTRAN or MPI to run on parallel servers was slow, difficult and crippled interactivity with the data," said Bryan Wong, lead author of the MIT paper. Star-P is an interactive parallel computing platform that enabled MIT scientists to use the models they already developed on PCs using their preferred desktop application, MATLAB(R). Star-P automatically parallelizes the models and runs them instantly and interactively on an 8-processor SGI(R) Altix(R) 350 server. "Star-P lets our scientists continue to work with their preferred tools, shielding them from the programming complexities of parallel systems. It automatically connects MATLAB to the server and parallelizes the application code on the fly," said Wong. For the initial research detailed in the technical paper, MIT scientists used Star-P's efficient parallelization system to obtain the thermodynamic properties of 1,3-butadiene -- a prevalent air toxin and petroleum combustion by-product -- for temperatures ranging from 50 to 500 degrees Kelvin. "Kinetic modeling of hydrocarbon molecules is critical to our understanding of the health and environmental risks they may pose," said Ilya Mirman, vice president of marketing at ISC. "The key to accelerating this kind of important research is to give our country's leading scientists faster, easier access to our best computational tools. Star-P brings supercomputing capabilities to the fingertips of a wide range of scientists who need parallel systems to do their work."