SCIENCE

Researchers solve electronic structures to explore biomass, supercapacitors, and more

Chemistry and materials science simulations made possible through Robert Harrison’s INCITE project have generated 70 scientific publications since 2008, many of which graced the covers of prestigious journals and dealt with topics from production of hydrogen for clean energy to development of graphene nanoribbons for power delivery. Collage courtesy Robert Harrison and Jason Smith

Enterprises from energy production to environmental cleanup depend on chemistry. Catalysts, which make chemical reactions more likely, contribute to nearly 20 percent of all industrial products. In the United States, three industries relying on catalysts—petroleum, chemicals, and pharmaceuticals—account for $1 trillion of the gross national product.

Catalysts are just one area of investigation for a multi-institutional team whose 70 publications in 3 years detail prodigious scientific output from the world’s fastest chemistry simulations. Its work integrates experiment, theory, and simulation to explore a continuum from chemistry to materials science. The researchers aim to understand, control, and design technologies needed for clean energy.

“Our long-term goal is enabling the design of new generations of clean and sustainable technologies to produce, transmit, and store energy,” said team leader Robert Harrison, a computational chemist at Oak Ridge National Laboratory (ORNL) and the University of Tennessee who directs the Joint Institute for Computational Sciences, a partnership between the two organizations. “Key elements are a fundamental understanding of chemical and electronic processes at the atomic scale and ultimately effectively transferring this knowledge in partnership with experiment into the hands of people and industries interested in the next steps of R&D.”

The research closely aligns with the Department of Energy’s (DOE’s) missions and its 20-year priority to support nanoscale science for new materials and processes. Two DOE Office of Science programs, Basic Energy Sciences and Advanced Scientific Computing Research, fund the work.

Through the Innovative and Novel Computational Impact on Theory and Experiment program, the researchers have been awarded more than 100 million processor hours since 2008. At the Oak Ridge Leadership Computing Facility, they calculate the electronic structures of large molecules and surfaces using scientific application codes called NWChem and MADNESS. The findings inform the development of processes, such as biomass conversion and fuel combustion, and products, such as batteries, fuel cells, and capacitors.

The electronic structure allows scientists to determine the positions and binding energies of atoms within molecules and responses to perturbations. Petascale computers speed complex calculations of molecular dynamics and quantum mechanics as substances undergo chemical transformation.

“Some of the largest calculations are only feasible on the leadership computers, not just because of speedy processors, but because of other architectural features—the amount of memory, the amount and speed of the disks, the speed and other characteristics of the interprocessor communication,” Harrison said.

A scientific paper by team member Edoardo Apra of ORNL was a 2009 Gordon Bell finalist after he and co-authors reported scaling NWChem to use most of Jaguar, ORNL’s Cray XT5 supercomputer, to calculate the electronic structure of water clusters, which is important in chemistry at interfaces and nucleation in the atmosphere. Because NWChem is a flagship application used throughout the community, getting it to run at petascale will have high impact. That was a special challenge, though, because the code employs distributed, shared memory instead of the message passing used by most codes. As a result of the team’s efforts, NWChem has now joined four other scientific applications sustaining more than a petaflop on Jaguar.

Further feats of computational science

Chemistry and materials science are critical for innovation. ORNL theorist Bobby Sumpter with lab colleagues Vincent Meunier and Jingsong Huang ran calculations of hundreds of teraflops on Jaguar to investigate a next-generation capacitor, a device that stores energy through charge separation at an electric double layer formed within porous materials. As transportation systems are electrified, fast charging of vehicles with such devices will be necessary. Iterating between simulation and experiment to reduce intermediate models, within 2 years Sumpter’s team and collaborators at Rice University arrived at a practical device for high-power energy delivery. Their supercapacitor stores—and quickly discharges—a thousand times more energy than a conventional capacitor. Several challenges remain to be overcome, however, before a commercial device can be developed and deployed.

“Because there’s very little chemical or material change, supercapacitors can be cycled millions of times without significant degradation. In contrast in a battery you’re physically moving considerable matter around and inducing chemical change, so it’s much harder to cycle more than a few thousand times,” Harrison said. “We’re familiar with the batteries in our laptops and cell phones dying after a year or two.”

While capacitors don’t seem chemically active, in fact a lot of chemistry is going on that’s relevant to energy storage. Indeed, Sumpter and collaborators achieved high energy densities in supercapacitors by coupling chemistry with nanomaterials—specifically, graphene nanoribbons structured at the billionth-of-a-meter scale of atoms. They considered the smallest unit of structure needed to achieve an energy-storage function and modeled a supercapacitor as a carbon pore. An ion and the shell of solvent surrounding it can move into the pore. But in materials with pores too small for the entire solvation shell to enter, some solvent gets stripped off. The ions align into a nanowire structure as they pack into the pore. The pore size and the ion’s tiny diameter determine the high capacitance.

Whereas Sumpter and collaborators simulate energy storage, other members of Harrison’s team address energy production. “When growing corn for conversion into ethanol, a lot of lignin and cellulosic material is left over,” Harrison said. “We could get energy out of it, and we could turn it into other useful chemicals—if only we had controllable, efficient processes.”

To improve those processes, computational chemist Ariana Beste and experimental chemist A.C. Buchanan, both of ORNL, explore thermochemical degradation of plant materials. They study how molecular structures influence networks of chemical reactions. The rate of a reaction depends on the height of energy barriers along paths between reactants and products and the fraction of molecules with enough energy to hurdle those barriers. One chemical reaction may lead to half a dozen products. Favoring a path that results in a specific product may necessitate understanding a hundred reaction paths.

Petascale simulations can quickly calculate the proportion of molecules with the requisites for a specific reaction—a herculean statistical challenge. Calculating which bonds between atoms in a molecule have the lowest energies, for example, reveals the optimal shape for a molecule to assume. That knowledge can speed design of processes faster than do trial and error or expert insight.

“Chemistry and materials sciences are sciences about the real world, which is complicated and messy, hard to characterize and control,” Harrison said. “We’re never going to be in a position to say we can replace chemistry with simulation. What we can say, confidently now, is that in increasingly large areas of our discipline, simulation and theory are equal and valued partners with experiment. Petascale computing is critical to accelerating scientific advancements.”