Titan Supercomputer Hunting for Better Biofuel in Massive Molecular Simulation

by Ali Albanese

Henry Ford once called ethanol “the fuel of the future.” That day may be moving closer, thanks to the Tesla GPU-accelerated Titan supercomputer at Oak Ridge National Laboratory.

A corn-derived biofuel, ethanol has the potential to be a widely used renewable transportation fuel, but its biomolecular structure has kept it from being economically viable to produce at mass scale. That’s where Oak Ridge scientist Jeremy Smith and Titan come in.

Smith and his team at the U.S. Department of Energy’s BioEnergy Science Center have used the Titan supercomputer to fastidiously produce some of the most immense and complex biomolecular simulations in history.

The main focus of their research is lignin, a vital component of plant cell walls that blocks enzymes from breaking down cellulose. Because cellulose breakdown is critical to converting plant materials into simple sugars, and eventually into biofuel, extracting the lignin is key to cost-effective ethanol production.

ORNL Titan supercomputer
The Titan supercomputer at Oak Ridge National Laboratory.

Super-Sized Simulations

Using a supercomputer-simulated 23.7 million-atom biomass system, Smith and his team could observe lignin’s binding properties on a large scale. The insights they gained helped them coordinate with other experiments to develop new chemical pretreatments and improve biofuel yield.

The results are promising: a 250,000-atom model consisting of lignin and an experimental solvent mix of tetrahydrofuran (THF) and water showed the solvent mix acting as a barrier between the water and the lignin. With a solvent like THF as a buffer, the lignin would be easier to remove during biofuel processing.

The model also demonstrated that lignin is selective in its binding preferences, favoring crystalline cellulose fibers. It’s the kind of detail that wouldn’t be possible to discern without Titan’s massive computer simulation.

Smith’s team was also able to observe how lignin operates with relatively smaller-scale, 100,000-atom simulation models. The team compared a wild and a genetically modified biomass system of lignin and hemicellulose (another major component of plant cell walls), gleaning insights into how the mutated lignin’s hydrophobic, or water-repelling, properties can boost biofuel yields. This knowledge can be used in future studies to make plants easier to deconstruct for biofuel production.