Fine-tuning experiments

11.01.10 -

by Vince Dixon

A Northwestern University researcher studying intermetallic compounds relies on high-performance computing to conduct atomic modeling prior to planning experiments.

When two or more metals are joined by soldering, intermetallic compounds often form at the interface between the base metal and solder. The solder joints are used in a variety of applications, ranging from food packaging to electronics packaging to plumbing. With numerous possibilities and many practical problems, testing the properties and finding good candidate compounds in solder joints for a particular application can be an expensive and lengthy process.

Gautam Ghosh, a research assistant professor in the department of material science and engineering at Northwestern University, wanted to predict the properties of intermetallics commonly found in solder joints without long and costly research. Usually, the mehanical properties of intermetallic phases are quite different from the base metal and bulk solder.

"That's the key advantage," Ghosh says. "If you can well define these experiments in advance, you can save a lot of time and resources."

Predicting the properties of soldered materials to save time and resources can benefit a number of research projects that use the materials, Ghosh says. In 2006, for instance, Ghosh and his collaborators received a grant from the Department of Energy's Office of Fossil Energy to use their research for the design and evaluation of ferritic steels. Using their computational design tools, they wanted to see how the steels would perform in high-temperature steam turbines. It's an effort that stems from several years of interdisciplinary collaboration between Ghosh's lab at Northwestern and other universities, member companies of the Semiconductor Research Corporation, and high-performance computing facilities.

NCSA was one of the supercomputing facilities Ghosh says has been important for the projects.

"NCSA has been useful to us for more than five years," he says. Ghosh had used NCSA's recently retired Mercury supercomputer and is currently using the Cobalt system.

Using Density-Functional Theory (DFT) codes, the researchers were able to predict various properties of metallic systems such as elastic constants, interfacial energy and bulk thermodynamics, Ghosh says. These properties are then related to potential behavior, allowing the researchers to predict certain intrinsic properties in solder interconnects, as an example.

"So, that's how we link our results, and try to predict our results with information behavior, elastic behavior, and so on," he says.

NCSA wasn't the only high-performance computing facility Ghosh used to design the experiments. The group used TeraGrid resources at the Pittsburgh Supercomputing Center, Oak Ridge National Laboratory, and the San Diego Supercomputer Center.

The use of high-performance computing plays a significant role in predicting the intermetallic properties by providing the researchers atomic scale modeling, but this is only one small step toward the lab's ultimate goal of adding length scale models to the project, says Ghosh.

With atomic models, the researchers make predictions using conditions that are, in theory, perfect. The end goal is to see how these predictions will work in the less-than-perfect real world, Ghosh says.

"What we have done so far is atomic-scale computations using ideal materials," he says. "But in reality, of course, materials are not ideal; there are many defects, for example."

The atomic modeling still helps researchers plan critical experiments, he says. Without them, performing experiments would be more extensive and time consuming. Instead of having to conduct dozens of experiments to test theoretical predictions, researchers can predict some of the results using the atomic models in order to fine-tune their approach.

Working with collaborators from the University of California at Davis and the University of Tennessee helped in better incorporating materials science and other fields into their research, Ghosh says. The collaboration also aided the team in gaining access to computational resources the Northwestern campus lacks.

"At Northwestern, there is no central parallel computing facility," Ghosh says. "So NCSA and other TeraGrid sites have been a tremendous help to us."

The help came just in time. While the study of solder materials has a long history at Northwestern, in recent years engineers at the university have begun to shift more focus to the study of intermetallics.

"That's because of the priority of the industry and the priority of the problems that need to be addressed," Ghosh says.

Using high-performance computing to predict the intermetallic properties of soldered materials may offer new areas of research in using intermetallics in interconnects.

This project was funded by the Semiconductor Research Corporation and the U.S. Department of Energy.