Materials science is on the verge of a revolution. Until recently, developing a new material has been a long and painstaking process: for example, lithium-ion batteries took 18 years to become a commercial product. The reason for this long timeline is that materials science research has largely been conducted by trial-and-error. So, when Duracell wanted to develop a new longer-lasting battery, they tried a different approach by working with Kristin Persson, current professor of materials science and engineering at UC Berkeley and Lawrence Berkeley Laboratory. Instead of testing many different materials to improve charge storage, they asked Persson for advice on which material to use. Using this data gathered from computation, not synthesis, the Duracell engineers saved years in developing their new battery. This work with Duracell was the impetus for the Materials Project. The goal of the Materials project is to help researchers better select which materials to synthesize to reduce the trial-and-error process.
“The Materials Project is an encyclopedia, where someone can come and find a new material to solve the problem they have,” explains Shyam Dwaraknath, a research scientist at LBL who works on the project. For each material, the Materials Project compiles a list of its properties, such as elasticity—a material’s stretchiness. They first turn to experimental studies describing the material property, but if none exist, they fill the gaps with their own computational simulations of the property. Researchers can then use the material properties they desire to select promising candidates for a given application. Rather than experimentally testing each potential material, they can use the Materials Project to identify materials that are predicted to display the ideal properties. This new, computational approach narrows the potential materials and occurs in a fraction of the time.
Unfortunately, many interesting materials have unknown properties. For instance, only about two percent of inorganic compounds have elastic constants available in the literature. One of the reasons this gap in information exists is that it is difficult to synthesize and test thousands of materials. To circumvent this issue, Persson and her team compute approximations for these properties using density functional theory (DFT), an approach that allows them to estimate the energy levels of a molecule or material without knowing its exact properties beforehand. Schrödinger’s equation allows scientists to solve for the exact energies of electrons in a material given its arrangement of atoms, but in practice, researchers use tools like DFT to easily compute an approximation of the material’s energy and infer various properties. For example, by repeatedly nudging the positions of atoms within the crystal and computing the resulting energies with each change, researchers can approximate the elasticity of the material. According to Persson, “It’s all about calculating the energy for different kinds of perturbations to the crystal. That’s how you get properties of the materials.” Obtaining these properties for thousands of materials is nearly impossible in a laboratory, but supercomputers make it a reality.
In addition to exploring current materials, the Materials Project determines properties of materials that have never been created in the lab by treating patterns in a crystal structure as templates. Each of these patterns is an arrangement of atoms, and substituting different atoms into the template gives rise to different—and sometimes completely new—materials. By searching through properties of these speculative materials, users can pin-down potentially useful materials without needing to iterate over hundreds of uninteresting ones in a lab. In particular, if a nonexistent material has desirable properties, a researcher might find it worthwhile to try to synthesize it.
The project has already had substantial impact. The Materials Project currently has over 170,000 registered users, making it the most popular publicly available database of materials in the world. Millions of data points are downloaded through their website every day, and early work has already helped spawn a commercial product, the Duracell Optimum battery. Yet there is still more work left. Now, Persson and her team are computing properties of non-crystalline materials and using artificial intelligence to suggest choices for new materials based on desired properties. While the Materials Project may have plenty of room to grow, it has already made an impact that suggests it will radically accelerate materials research as we know.
Alok Tripathy is a graduate student in electrical engineering and computer science.
This article is part of the Spring 2021 issue.
Notice something wrong?
Please report it here.