Image courtesy of Chris Wolverton
This image is a network graph—called a "minimum spanning tree"—showing the 7,410 DFT-predicted stable compounds from the Open Quantum Materials Database (OQMD) at the time the JOM article was completed. Since then, the number of stable compounds predicted in OQMD has risen to about 18,000 (out of 300,000 total compounds in the database). The tie lines connect compounds that are in stable equilibrium with one another. The plot was generated with the open source Gephi program (http://gephi.org).
Materials lie at the very crux of technological innovation. It's no accident that the birthplace of the personal computer is known as "Silicon Valley." As has sometimes been pointed out, ever since the transition from the Stone Age to the Ages of Bronze and Iron, new materials have defined new technologies and whole new ways of life. (To call our own era the Age of Silicon would not be so farfetched.)
Over the years, America has been prolific in developing new materials—including silicon semiconductors and countless others—and DOE and its predecessor agencies have played a big role in supporting basic research on materials at both universities and national laboratories. Today, with global competition challenging U.S. economic leadership as never before, there is an acute awareness among policymakers that America's future competitiveness will depend on a continued capacity to innovate in materials. And in fact there is a strong sense that the pace of innovation needs to be accelerated. New materials have traditionally taken from 10 to 20 years to get from the lab bench to the marketplace. In today's competitive world, that pace looks too slow.
So a new government initiative begun in 2011—dubbed the Materials Genome Initiative, or MGI—is seeking to mobilize U.S. scientists to find new ways to step up the pace of materials discovery and innovation. The name, which uses the term "genome" metaphorically, appears at least partly to have been chosen to evoke the memory of another government initiative, the Human Genome Project, or HGP, whose transformational impact on both science and the economy is remembered as a huge success.
Of course, the question arises, what makes anybody think that materials discovery and innovation can suddenly be sped up? The answer in a word: supercomputers. High-end computation, taking advantage of faster and faster machines, is transforming not just materials research but nearly every corner of science. That is because high-end computation enables researchers to achieve within hours or days in virtual space what might take years or even a lifetime in a physical laboratory—or simply couldn't be accomplished at all in the physical world.
Recent work out of a DOE Energy Frontier Research Center (EFRC) led by Argonne National Laboratory, with Northwestern University as a major partner, shows how the MGI is beginning to bear fruit through just such use of computation. The research was led by Chris Wolverton, Professor of Materials Science and Engineering at Northwestern, and included fellow Northwestern researchers James E. Saal, Scott Kirklin, Muratahan Aykol, and Bryce Meredig. It was reported in JOM, the journal of The Minerals, Metals, and Materials Society.
Researchers in different fields are using high-end computation in a variety of different ways—for modeling and simulation, data mining, virtual prototyping. Here researchers have used high-end computation for what are known as ab initio (Latin, "from the beginning") calculations. They have performed systematic analyses of thousands of known—and thousands more putative or imagined—chemical compounds based on first principles, illuminating key properties in the process. They have established a database of the results of these analyses and are providing search and data mining tools for researchers to access the data. In an express response to a call in the MGI, the researchers have made the full database and search capabilities—which they call the Open Quantum Materials Database (OQMD)—publicly accessible on the web (please see link below).
Traditionally, a lot of materials discovery has been by trial and error. OQMD moves much of this trial and error process from the lab bench to the computer, where it can be radically accelerated.”
The purpose of the effort is to enable researchers to identify candidate materials—both existing and new—for specific applications by screening them computationally for various properties before they are synthesized or tested in the laboratory. Traditionally, a lot of materials discovery has been by trial and error. OQMD moves much of this trial and error process from the lab bench to the computer, where it can be radically accelerated.
The JOM article provides several examples of applications where the researchers using OQMD were able to identify plausible candidate compounds—both existing and new—and thereby rapidly narrow the search and quicken the trial-and-error process. Among the examples are materials for the anode of lithium ion batteries, reaction compounds for a lithium-air battery, coatings for the cathode of a lithium-ion battery, and materials for new magnesium alloys (for increasing energy efficiencies by reducing vehicle weight, among other potential applications). Typically, the researchers were able to define a screen or filter that captured existing compounds already effectively used for the application along with a handful of promising alternatives, including novel compounds. The OQMD searches proved to be big potential time-savers. In the case of the lithium-air battery reaction, for example, the search narrowed the potential candidates from 255 to 10—leaving just a handful of materials to be tested in the laboratory.
The underlying physics of the project goes back to some Nobel Prize-winning research. The approach goes by the name of "density functional theory" (DFT). It rests on a proof, put forward in the mid-1960s by physicists Pierre Hohenberg and Walter Kohn, that the electron density—a quantity that is relatively easy to compute, whether for an atom, molecule, or compound—has a predictable relationship to the total energy, which, as it happens, is extremely difficult to compute directly, using Schrödinger's equation, for any system consisting of more than a very few atoms. Though DFT was first proposed in the mid-1960s, it only came to be practical after some years of mathematical refinement and, crucially, the growth in the power of computers. By the 1990s it was a cornerstone of computational chemistry. Kohn's Nobel Prize was awarded in 1998.
DFT makes it relatively easy to compute the energy of a compound and compare it to others or to the pure materials from which the compound is formed. The one with the lowest energy is stable and thus the preferred structure. By computing the energy of thousands upon thousands of different configurations of groups of atoms—the database is currently at about 300,000 and growing—one can determine which compounds are likely to be stable. DFT can also be used to predict certain properties of crystalline solids, including, for example, the geometry of the lattice, the level of magnetism, and the energy required for the crystal to form.
With today's extremely rapid expansion of computational capability, DFT has truly come into its own. Wolverton and his colleagues call their approach "High-Throughput Density Functional Theory," referring to the large quantities of data that now can be processed at speed using today's high-performance systems. In their research the team made use of both Northwestern University's high-performance computer Quest, and the National Energy Research Scientific Computing Center (NERSC), a DOE Office of Science user facility at Lawrence Berkeley National Laboratory. Support for the research also came from the DOE Office of Science, primarily through the Center for Electrical Energy Storage EFRC, and from several other sources, including Dow Chemical Company, the Ford-Boeing-Northwestern Alliance, and the Department of Defense.
Whether MGI could have an impact comparable to that of the 10-year $3.8 billion HGP remains to be seen. But it is clearly inspiring researchers to accelerate discovery and innovation by marrying computation and experimentation in new ways.
—Patrick Glynn, DOE Office of Science, Patrick.Glynn@science.doe.gov
James E. Saal, Scott Kirklin, Muratahan Aykol, Bryce Meredig, and C. Wolverton, "Materials Design and Discovery with High-Throughput Density Functional Theory: The Open Quantum Materials Database (OQMD)," JOM 65, 1501 (2013).
Research: DOE Office of Science, Office of Basic Energy Sciences; Dow Chemical Company; Ford-Boeing-Northwestern Alliance; Department of Defense
User Facility (NERSC): DOE Office of Science, Office of Advanced Scientific Computing Research
Open Quantum Materials Database
Center for Electrical Energy Storage