FEATURES & COMMENTARY LIVEwire

Dallas, Texas — As scientists study interactions at incredibly small levels, the implications of their work grow larger and larger.

Imagine an automobile air bag system that can calculate not only when but with how much force, to deploy an airbag based on the passenger’s size and weight. Picture a home water filter system that monitors pollutants as they pass through, sensing when levels rise too high and automatically adjusting the filtration process. Or, think about working in an office that could sense the your movements and reconfigure the office technology accordingly – for example, from shared video conference to private work mode.

Welcome to the world of microelectromechanical systems (MEMS), an emerging field that uses microfabrication techniques to bring together electronic circuitry and three-dimensional structures and devices such as sensors and acuators onto silicon chips. In these systems that are measured in the millionths (micrometers) and billionths of a meter (nanometers), a working gear, sensor, or filter may be no larger than a grain of sand. More importantly, the behaviors of particles and individual electrons at this scale don’t follow the rules of classical physics. Engineers and scientists are working to document the basic behaviors of particles and electrons on the micro- and nanoscales. Their work could bring about reliable, low-cost integrated systems-on-a-chip that are “smart” enough to sense and respond to the needs of the user.

“Our ultimate goal is to create embedded systems that result in smart surfaces,” says Umberto Ravaioli, a University of Illinois electrical and computer engineering professor and a researcher with the Alliance Nanodevices team. Ravaioli says scientists are “just beginning to get their feet wet” in MEMS. He is one of three U of I researchers who are dipping their toes into the field by using the Alliance’s high performance computing systems at NCSA to conduct research into the behavior of air particles as they flow through microfilters no more than the diameter of a human hair. Narayan Aluru, an assistant professor in the U of I department of general engineering and a researcher in the Beckman Institute for Advanced Science and Technology, and graduate research assistant Ozgur Aktas, are also part of the research project.

The research team is looking at how very small airborne particles flow through the microfilter’s elements, tiny holes that are no larger than 1 micrometer. These miniature filters are used to sift very small particles such as spores and bacteria. The entire filter array is usually only a few micrometers in diameter, or about the size of a quarter. As a micromechanical component of MEMS, these filter elements are often micromachined onto silicon. For example, a hand-held device to detect gas leaks could include not only a gas spectrometer small enough to fit on a chip but also a microfilter that would sift out extraneous particles such as dust. Including a microfilter on the chip would effectively purify the gas and allow the spectrometer to detect only the components it was designed to detect.

“At this point we need to study the behavior of these particles under various conditions because we don’t know much about the behavior of gases at this level,” says Aluru. “Computer simulation is the closest we can get to observing what happens in the real world.”

Aluru used funding from the NCSA Faculty Fellows program as seed money for the current research project. The Faculty Fellows program awards grants to faculty on the U of I’s Urbana-Champaign campus for research projects that could benefit from the use of NCSA or Alliance computing resources. The researchers are running simulations on NCSA’s NT supercluster and SGI Origin2000 supercomputer using Direct Simulation Monte Carlo (DSMC) techniques to simulate the behavior of air particles in filters ranging in size from .05 to 1 micrometer. The DSMC method samples a significant number of the particles as they flow in, out, and around the intake area of the filter element. From this random sample, the researchers draw general conclusions about the behavior of the particles.

The team’s first simulations involved about 5 million molecules flowing through the filter element. Each simulation-one on the NT supercluster and one on the Origin2000-used 64 processors for about 14 hours. An accurate 3D simulation of all flow features of a microfilter element would require about 300 million molecules, something that is practically impossible even on the best of today’s supercomputers. To address this problem, the team is developing new multiscale methods of computing and simulating their data.

The team’s simulations track a wide range of conditions such as air pressure and temperature at different flow rates as well as differences in pressure at the filter’s intake and output points. The simulations follow each of the millions of molecules on their journey through the microfilter, noting when they flow smoothly through the filter, when they hit the walls of the filter, and when they collide with each other. Understanding these basic behaviors of gases at this level, says Aluru, can help answer some elementary design questions, such as the optimal shape of a microfilter or how much pressure a tiny filter can take before it bursts.

“There are some basic engineering questions that need to answered,” notes Aluru. “We need to know about the effects of pressure differences between intake and output, the effects of more or fewer [filter] holes, how the roughness of the surfaces effects flow, or what happens when you reduce the interactions among gases.”

Results so far show that the behavior of filter elements is not governed by classical models of fluid transport. In addition, surface conditions in the filter elements-such as roughness and how particles interact with the surface of the filter-play important roles in determining the behavior of particles at very small scales. Additional 3D simulations will provide even more insight into the behavior of particles at this scale and will lay the foundations for developing nanoscale filter elements, Aluru predicts.

“We’ve learned a lot, but there’s a lot more that needs to be done,” says Ravaioli. He hopes that in the not-too-distant future, simulations of the workings of entire microelectromechanical systems will be possible-simulations that could require as many as 40,000 processors on clusters or Origin2000 systems.

“This is an emerging discipline that will require an enormous amount of compute power,” he says. “It will push the development of terascale computing systems.”

This research is supported by the Defense Advanced Research Projects Agency, the NCSA Faculty Fellows Program, and a University of Illinois Computer Science and Engineering Fellowship.

============================================================