"The Little Engine," Feature Article, March 2007

FEATURE FOCUS: Microsystems

the little engine

Can a gas turbine with tolerances smaller than a wavelength of light make power more portable than ever?

by Michael Abrams, Contributing Editor

batteries are too big. And they don't last long enough—just ask any soldier, laptop user, or TV cameraman. But Alan Epstein, a professor of aeronautics and astronautics at the Massachusetts Institute of Technology, hopes to change all that with a gas turbine engine made of silicon. It's no larger than a quarter and can be stamped out a hundred at a time.

Epstein and his colleagues have been working on the little engine for more than a decade now, and they may currently be just months away from an actual working model. It's hard to tell exactly, because, unlike the fixes that might be needed to nudge a full-size turbine to readiness, every change Epstein's team makes means starting over and building the engine again.

"That's the big difference between something built in silicon, and something built conventionally," Epstein said. "If it's conventional, and you decide something's too big, you take it apart, take it down to the machine shop, then reassemble it. With our engine, once you've built it, it's one solid piece of silicon, and to make a change you have to start from scratch."

Microengines for microprocessors: These tiny silicon gas-turbine engines may soon power laptops or cell phones. And, they'll do so efficiently.

However long and difficult the design cycle, a big surprise for Epstein was discovering how similar the overall concepts of a microengine were to a turbine of any size. "We thought we'd have problems that were very different from a large engine, but in retrospect, we haven't. Our solutions are designed differently, but the challenges are the same: bearings and rotor dynamics," Epstein said.

Contrary to previous analysis, the fluid mechanics at the size Epstein hoped to build his engine turned out to be the same as those of larger engines. As long as the passages made for gas flow are larger than a micrometer in diameter, molecular kinetics are not an issue. The size of the tubes is not so small that at the molecular level the behavior of the fluid against the passage walls changes.

That said, the size of the engine does alter the design, of course—mostly thanks to the limited way tiny things are built in silicon. Whereas a larger engine might first be designed for efficiency—with the question of how to actually manufacture it put off till later—the unique problems of manufacturing in the minuscule dictate the design from the get-go.

Fine Etchings

To make whole sheets of the little turbines all at once, they are built with nine etched and bonded silicon wafers (earlier versions used only six). The virgin silicon is first coated with a photoresist, then a design pattern is applied on top. Next, the wafer is developed and baked. The silicon that remains exposed is then etched, either chemically or with a plasma. To protect the resulting vertical walls from being worn away, they are dusted with a Teflon-like polymer. (The area covered by the pattern is actually etched as well, but as the rest of the silicon is removed somewhere between 50 and 100 times faster than the pattern, the desired depth is achieved.) By repeating this process, a single wafer can have several layers. Smooth slopes may someday be achieved with a gray- scale pattern being developed at MIT's partner, the University of Maryland.

The rotor and its airfoils are carved out of a single wafer. Additional plumbing and bearings are etched onto the wafers that are to sandwich the rotor. All the layers must then be bonded together. Silicon bonds well to silicon, it turns out, and the bonded areas are just as strong as the material itself—but only if the surfaces are kept perfectly clean. A dust particle no bigger than a millionth of a meter in diameter can keep an area the size of the engine itself from bonding.

Although it would be feasible to place a separate rotor into the middle of a silicon engine, the cost and time required for such a procedure would be prohibitive—making the engines impossible to produce cheaply by the hundreds out of a single silicon sheet. Instead, the rotor is made entirely out of one of the wafer layers, but it cannot be completely freed during etching or this most crucial element may fall out during the rest of the manufacturing process. To keep from losing it, Epstein's team keeps it attached until the very end, either with a glue that can later be dissolved or with thin silicon tabs that are easily broken.

This cross-section is of an earlier concept using only five layers. The center wafer contains the etched rotor (disc and airfoils) and the rest of the sandwich consists primarily of bearings and plumbing.

Combustion occurs just outside the rotor, at the same wafer level, spinning it by pushing on its airfoils from the outside. At more than a million rpm the heat produced by the spinning rotor threatens to actually weaken the silicon, so cooling becomes a major issue. To pump out the heat more quickly, the shaft that would normally be in the center of the rotor is removed. A side benefit of the high rpms is that to human ears the turbine is silent. Electricity will be produced with either a tiny magnetic generator, or an electrostatic induction machine.

"To date, these have been driven by micro air turbines for test purposes, rather than the micro gas turbine, which has yet to produce positive mechanical power," Epstein said. With the air turbine the magnetic generator has been shown to produce 10 watts of power.

Although the turbine size is not small enough to change the behavior of fluids, it is small enough to make any fine tuning of the plumbing difficult. On large turbines, for instance, changes in fluid density are handled with tapered passages. Such tubing is currently impossible on an engine of this size, although changing the rotor's airfoil thickness can help the problem somewhat. There's also no way to make tubes with gentle curves—passages are necessarily either etched straight down through a wafer, or across it—so the plumbing has to change directions at right angles. Both limitations reduce the overall efficiency.

From top to bottom, left: (1) A magnetic generator, 4 mm in diameter; it's almost 60 percent efficient. (2) A 6 mm diameter turbine nested neatly within the cumbustor. (3) The airfoils on this silicon wafer have thick trailing edges to make up for an inability to taper fluid paths. (4) A diamond saw will separate turbines along the lines. Empty white circles test how the material responds to processing, and squares within the squares contain MIT's logo.

Small losses like these can add up and it's been shown that in general the smaller the engine, the lower the efficiency. (Epstein, however, points out that this may have more to do with the funding available for small engines than it does with fluid flow.)

According to Epstein, the fuel source could be packaged with the engine or come as a cartridge like a cigarette lighter. "Do you refuel it like you refuel a lighter?" he asked. "Do you sell cartridges? What you rapidly realize is that it's all fuel. How you choose to market and package it is a market question."

In order to have any longevity with the high rotor speed of an engine of this size, the bearings must be low friction.

Epstein's group considered magnetic bearings early on, but found that in addition to the manufacturing difficulties, the magnetic materials had too low a Curie point and would not stay magnetic at the temperatures at which the engine would operate. Instead, they chose pressurized gas bearings, which conveniently can hold more weight relative to their size as they get smaller. Thrust bearings with spiral grooves and holes in their centers are self-pumping and keep the rotor free and in the right position.

Bearing the Loads

"You can indeed make million-rpm air bearing systems out of silicon and have them run reliably," Epstein said. But the bearings on such an engine have to be able to withstand not only the forces going on within, but also the sudden acceleration that might occur when, say, a cell phone is yanked off a table—or dropped on a sidewalk.

The concern is even greater for Epstein's lab since his prototypes take months to put together and are assembled one at a time. "These things are fragile, and if someone drops the wafers—it's happened at Intel. People drop things that are worth tens of millions of dollars."

As for the primary materials in use, with the etching process, the options were few. "Our choices were silicon, silicon, silicon, so we chose silicon," said Epstein. However restricted they were in their materials selection, silicon turns out not to be too bad: It can go to higher temperatures than the materials used in larger engines, and is stronger, too. Silicon nitride and silicon carbide would work well in larger turbines if it weren't for the fact that they are difficult to manufacture in large sizes without introducing flaws.

While there is clearly plenty of room for improving efficiency, the microengine may very well end up as the only real way to power, say, a laptop, an iPod, or a soldier's thermal weapon sight, to say nothing of a palm-size plane. In terms of power per pound, the little engine will easily beat batteries with an output of somewhere between 50 and 100 watts and a 100:1 thrust ratio. Overall it will perform as well as the gas turbines made in the 1940s.

So what, then, is the holdup? "We're at the stage where we chose to demonstrate each part separately. All of them work as individual devices," Epstein said. "It's getting them all to work on the same day and at the same place that's the challenge."