A-Z Index | Directory | Careers

Novel Tools and a Pharmaceutical Screening Strategy to Capture CO²

December 6, 2012

Contact: Linda Vu, lvu@lbl.gov, +1 510 495 2402

Today, crystalline porous materials, like zeolites and metal organic frameworks (MOFs), are widely used in industry to purify water and separate gases, among other things. But scientists believe that these structures have the potential to do a lot more—like capturing carbon dioxide (CO2) from the flue-gas emissions of coal-burning power plants before it reaches the atmosphere and contributes to global warming.

Until recently, one of the major challenges to using these materials for CO2 capture was identifying the right porous structures to effectively do the job. But novel tools developed by Lawrence Berkeley National Laboratory (Berkeley Lab) computational researchers, combined with an informatics screening strategy inspired by the pharmaceutical industry, is making this search a lot easier.

“In the category of zeolites alone, there are about 200 known structures and 2.5 million structures predicted by computational methods,” says Richard Luis Martin, a postdoctoral researcher in Berkeley Lab’s Computational Research Division (CRD) and lead author of a paper recently featured on the cover of ChemPhysChem.

He notes that researchers previously used brute-force techniques—like exhaustively running molecular simulations of individual structures—to search large databases and see whether any materials would be successful for a desired application, like capturing CO2. Although these methods showed results, they also proved to be computationally expensive and time-consuming.

“The computational cost of these calculations prohibited the characterization of databases containing millions of structures, and because the structures that exhibit desired properties constitute only a small fraction of predicted materials, the brute-force screening approach can involve many wasteful calculations, ” says Martin.

So to effectively and efficiently screen massive databases for materials that will successfully trap CO2 molecules, Martin and his colleagues combined their novel computational tools with a screening strategy that has successfully been employed in drug discovery.

“Instead of using a brute-force approach and examining every single porous material in a database, we use our Voronoi hologram tool within Zeo++ to first select a small sample of materials within a database, each with very different pore structures,” says Martin. “Our method is based on a property principle, similar to the one successfully employed in drug discovery, which states that similar chemical structures have similar properties: a simple yet powerful concept that leads to efficient chemical space exploration.” 

Because researchers deal with only a simple representation of the material’s empty space, rather than all the atoms in the structure, Martin notes that Zeo++ can characterize structures much faster using far less computing power. In the paper, Martin and his colleagues describe how they used this tool to identify a sample of 130 materials with widely differing substructures from a database of about 140,000 materials. Once they identified a sample of 130 materials, the researchers ran molecular simulations to better determine which materials were good candidates for trapping CO2 molecules. Then they looked for commonalities in the substructures.

“We found that the best materials in this small subset had particular features in common, small regions within the pore network that were just the right shape to envelope and bind a single CO2 molecule; we called these regions ‘sweet spots’,” says Martin.

The team then used their shape-matching algorithm to scan the entire 140,000 structure database and looked at the 500 materials found to have the best shape matches. Within these 500 materials, the team identified 327 structures that had the sweet spots they were looking for, and which would be extremely effective at capturing CO2 molecules.

“We already knew the performance of each material in the 140,000 structure database from a previous study where we used supercomputers to perform molecular simulations on each structure,” says Martin. “Because of this, we know that our algorithms are about 65 times more efficient at selecting good candidates.”

When they applied this technique to a database containing one million zeolite structures, they found about 30,000 candidates for effectively capturing CO2 molecules.

“Our algorithms allow us to discover the majority of high-CO2 absorbing zeolites by characterizing about one percent of a database. Using these methods, we achieve shorter discovery time and cost, ensure that valuable computer resources are used to characterize only important materials, and can screen much larger databases than ever before,” says Maciej Haranczyk, a scientist in Berkeley Lab’s CRD and a principal investigator on the project.

Haranczyk notes that the algorithms and screening strategy allow researchers to scan databases with several millions of structures, something that was simply impossible to do just months ago. Using the shape-matching tool, the team searched a database containing more than 1 million structures in less than eight and a half days, using a single desktop machine (10 CPUs in parallel). This translates to about seven seconds to characterize one structure on a single CPU.

“By comparison, performing an actual molecular simulation of a single structure using a single CPU on a desktop machine, can take anywhere from 30 minutes to several hours,” says Martin.

In addition to Martin and Haranczyk, other authors on the paper include: Thomas Willems and Jihan Kim, also of Berkeley Lab, with Li-Chang Lin, Joseph Swisher and Berend Smit, of the Department of Chemical and Biological Engineering at the University of California, Berkeley. 

Both the Voronoi hologram and the shape-matching algorithm are part of the Zeo++ suite of tools.

About Computing Sciences at Berkeley Lab

High performance computing plays a critical role in scientific discovery. Researchers increasingly rely on advances in computer science, mathematics, computational science, data science, and large-scale computing and networking to increase our understanding of ourselves, our planet, and our universe. Berkeley Lab’s Computing Sciences Area researches, develops, and deploys new foundations, tools, and technologies to meet these needs and to advance research across a broad range of scientific disciplines.