The idea of using lithium-air batteries to power electric cars has been around since the 1990s, but the last few years has seen a big upsurge in interest for this technology. Oil supply concerns, global climate change, and a general discontent with the internal combustion engine are driving the search for more robust batteries, and lithium-air is at the top of the list.
In fact, lithium-air has been described as “the holy grail of battery material” by some in the industry. Unlike lithium-ion batteries, which generally can deliver 50 to 100 miles per charge, lithium-air technology promises up to 500 miles — about the same as gasoline-powered cars. If perfected, the technology will make widespread use of electric vehicles possible.
Lithium-air batteries have a natural advantage in weight and energy density compared to their lithium-ion cousins. Since atmospheric oxygen is used to power the reaction (rather than a lithium-metal complex in the ion version), lithium-air represents a much lighter solution. And since the battery takes advantage of the considerable energy released from the lithium-oxygen reaction, the energy density is much higher. On paper, the maximum energy density of a lithium-air battery is 5,000 watt-hours per kilogram. That works out to a 10-fold advantage when compared with the lithium-ion hardware under the hood today.
But lithium-ion technology has a big head start, and those batteries are already slated to go into the new GM Volt and Tesla electric cars. Further development of this technology should improve performance, but the physics of lithium-ion technology will limit how far it can go. Lithium-air batteries, on the other hand, are still in the research stage and may not see the light of day for another five to 10 years, perhaps even 20 years.
Given the array of energy problems the world is struggling with, the US Department of Energy (DOE) and companies like IBM would like to speed up that timetable considerably. To that end, a DOE INCITE award has allocated 24 million hours of elite supercomputing time to support lithium-air battery design. Argonne National Lab’s half petaflop Blue Gene/P machine will supply 12 million of those hours, with the other 12 million coming from Oak Ridge National Laboratory’s 1.7 petaflop “Jaguar” Cray XT. Researchers from Argonne, Oak Ridge, Pacific Northwest National Laboratory, and IBM are all involved in the computational work, in a project nickamed Bat 500 (for batteries that provide a 500-mile range in electric vehicles).
According to Larry Curtiss, a senior computational chemist working on the project at Argonne, the support work for lithium-air technology has been going on for around six months at the lab, using both mid-range HPC clusters and the “Intrepid”* Blue Gene/P super. The computational codes are used to help design the battery components, including the internal membranes and electrolyte material, as well as the electrode compounds. The scale of these simulations runs from the atomic level up to larger scale modeling of the battery’s operation.
The basic reaction is rather simple. Lithium reacts with oxygen from the air forming lithium peroxide, releasing electrons in the process. But the reaction cannot be one of uncontrolled oxidation. (Elemental lithium is too reactive in both oxygen and water to let it just roam unescorted.) Rather, the lithium at the anode has to be transported via an electrolyte to meet up with the oxygen in solution. Recharging the battery involves decomposing the lithium peroxide back to lithium and oxygen. Since the peroxide form of lithium is quite stable, this reaction requires a catalyst at the cathode end to make it practical at normal temperatures.
As simple as this sounds, a lot of science is required to design the optimal materials that make it all possible. For example, the electrolyte and the lithium anode must both be stable so the reaction takes place quickly and the lithium metal eventually finds its way back to the anode. Also, the catalyst must be efficient at breaking up the lithium peroxide.
“That’s one of the big challenges — to find a catalyst material to break the lithium-oxygen bond,” says Curtiss. His computational work is focused in that area, and he uses density functional theory methods to model the bond-breaking and bond-making in these catalytic reactions. Those methods simulate electron-level effects and can span thousands of atoms. The next level up involves dynamic simulations that can track millions of atoms.
While these codes are all very compute-intensive, at the smaller problem sizes, they can be run on standard vanilla clusters. But when the researchers need to run more realistic simulations, they turn to big supers. Curtiss says the largest simulations can keep 30 thousand Blue Gene/P cores busy for up to 48 hours. Of course, the researchers would love to increase their model sizes even further, not to mention speed up execution times. For that, the researchers at Argonne will have to wait for maybe a couple more years — when the first Blue Gene/Q machine shows up.
*Editor’s Note: The original version of this article incorrectly referred to Argonne’s BlueGene/P supercomputer as “Dawn” until a reader comment pointed out the correct name is “Intrepid.” For the record, “Dawn” is the name of the BlueGene/P system housed at Lawrence Livermore National Laboratory.