In pioneering work that has important implications for the field of microscopic machines and could answer basic questions of materials science, MIT researchers are developing extremely tiny and sensitive test systems to measure crack growth in silicon and other materials.
The researchers, led by Professor Stuart B. Brown of materials science and engineering, have already created a system that can measure crack growth in single-crystal silicon with a resolution of nanometers. This resolution permits detection of average rates of crack growth as slow as 10-15 meters per second, which corresponds to a rate of one atomic spacing every two days.
"That's three orders of magnitude better than any other technique we know of," Professor Brown said.
Such sensitivity "will allow us to probe material behavior that hasn't been accessible till now," Professor Brown said. For example, he said, "there are many questions about how cracks start and grow in brittle materials."
The work is also important to the emerging technology of micromachines. A crack that grows as slowly as those tracked by Professor Brown and colleagues may seem inconsequential-most everyday products we're familiar with would fail for other reasons long before they'd fail because of such cracking-but think about the implications for machines so small that most can't be seen with the naked eye.
At that size "there's less tolerance for defects," said Professor Brown. "Even very small cracks can cause the device to fail."
Such micromachines are already on the market. For example, they are used in cars as tiny sensors that can detect such things as pressure and chemical vapors. Current applications, however, are only an appetizer of things to come. Scientists envision microsensors aboard automated submarines that would measure water quality without having to take large water samples, and gnat-sized robots powered by micromotors that could be used to explore Mars.
Professor Brown had no direct experience with micromachines when graduate student John A. Connally asked him to be his advisor on the subject in 1988, so at first he turned him down.
But Connally came back and asked again, and this time Dr. Brown said yes, on the condition that the research be in the mechanical behavior of the materials used to fabricate micromachines.
In the end they settled on the mechanical reliability of micromachines. But when the two did a literature search they made an interesting find: virtually no work had been done on the subject. "I found that striking," Professor Brown said, "because one of the advantages of these devices is they're supposed to last about 10 years in fairly aggressive environments, yet no one's sure they can last that long."
The researchers also found that it wasn't clear whether cracks would grow in silicon-the material most micromachines are made of-that was exposed to the humidity of an average room. "The conclusion was that cracks wouldn't grow, because no one could detect such growth," Professor Brown said. But again, since even very slow crack growth rates could catastrophically affect micromachines, a conclusive answer was important to Connally and Brown.
So the two set about designing a test system that could measure ultra-slow crack growth in silicon. The result? After an interdisciplinary effort that included expertise from MIT researchers in electronics, materials science, solid mechanics and electronic fabrication, "we built what I think is the world's smallest system to detect crack growth in a material," Professor Brown said. The critical parts of the system span a distance only about 400 millionths of a meter wide. That's roughly 1/60th the width of a standard 29-cent US postage stamp.
RESULTS AND IMPLICATIONS
Last year, Connally and Brown reported in Science magazine that the system had indeed detected crack growth in single-crystal silicon exposed to moisture. "Our results were unambiguous," Professor Brown said. Over time, the crack growth rate they measured was only 10-13 meters per second.
These results have a number of implications for micromachines. For one, Professor Brown said, "they indicate how these devices may fail, and what the operating limits are." He noted that the results also indicate that "there may be modes of failure unique to these devices that people should be aware of." One obvious conclusion from the work: "engineers should design sensors [and other micromachines] so they're hermetically sealed," Professor Brown said.
Professor Brown notes that similar systems could be made to detect crack growth in other materials like tungsten and silicon nitride, which future generations of micromachines might be made of. Professor Brown and Gary L. Povirk, a postdoctoral associate in materials science, are currently working on a system to detect crack growth in polycrystalline silicon. (Dr. Connally received his PhD last year.)
HOW DOES IT WORK?
The system Connally and Brown designed was created at the Charles Stark Draper Lab (Draper also funded the project). It is composed of a tiny table-tennis paddle that extends two thirds of the way over a square hole. The paddle is part of the surrounding material (in this case single-crystal silicon); the cavity beneath it is etched away using semiconductor technology. Sitting atop-but not touching-the cavity and paddle are two skinny rectangular gold electrodes. A dab of gold on the end of the paddle serves as a weight.
Finally, at the base of the paddle the scientists introduce a small crack. (They could have tested the structure without introducing a crack, "but then crack formation is dependent on the properties of that particular structure," Professor Brown said, which introduces a number of other variables. "We wanted to have a well-defined defect.")
The system works by vibrating the paddle via the electrodes, which creates stresses at the base of the paddle. "We can't wait 10 years to see if the crack will grow, so we do accelerated testing," Professor Brown said. The scientists put the system under conditions that are more severe than most micromachines would necessarily encounter in use.
Every structure will vibrate at a certain frequency-the resonant frequency. And if the crack in the base of the paddle grows, this resonant frequency will change.
To determine crack growth the researchers employ a mathematical model of the structure with different crack lengths and their corresponding frequencies. "So we measure the resonant frequency of the actual specimen [at a given point in time], then use the model to tell us, given that frequency, how long the crack is," Professor Brown said.
In the end, then, John Connally's persistence paid off. "We addressed some basic questions in materials science and some critical issues important to the micromachine community," Professor Brown said. "John certainly earned his degree."
A version of this article appeared in the March 10, 1993 issue of MIT Tech Talk (Volume 37, Number 25).