The entire device—complete with an integrated electric generator—is expected to weigh in at just 1 gram. According to lab director Alan Epstein, a prototype silicon microturbine produced using semiconductor-type microfabrication methods may be operating by the turn of the century. If that initial effort meets success, the researchers plan to use similar lithographic techniques to construct another radial inflow turbine engine from silicon carbide, a refractory ceramic material.

The researchers have already manufactured a 4-millimeter-diameter radial inflow turbine wheel from silicon using deep reactive ion etching—a relatively new microfabrication method, according to Epstein. With minor changes in the airfoil shapes, the same component will function as a centrifugal compressor wheel as well. Work is also progressing on a suitable combustor unit, he said.

Calculations indicate that this kind of turbine driving an electrostatic induction generator of similar size could supply tens of watts of continuous electric power. If it were made of a temperature-resistant material like silicon carbide, a complete gas-turbine generator system—with a volume less than 1 cubic centimeter—could deliver as much as 50 watts of electric power or about 0.2 newton of direct thrust.

The benefits of such a high-risk technical development effort could be considerable, according to the MIT team. Stamped out in large quantities like semiconductor chips, the potentially low-cost micro-heat engines could see widespread application as mobile power sources, propulsion engines for small unmanned air vehicles, aerodynamic boundary-layer and circulation control (wing blowing), and coolers for both electronics and individuals.

Perhaps the most enticing application would be for portable power production, a niche now filled by electrical generator sets or batteries. The energy density of liquid hydrocarbon fuel is 20 to 30 times that of the best battery technology, so the possibility exists that these power sources could shrink proportionally. The tiny turbines would likely consume less than 10 grams per hour of hydrogen fuel, according to the MIT team's estimates. The fuel supply, an exhaust system, and presumably a thermally insulating containment vessel would add to the unit's size and weight, of course. If additional power were needed, several microturbines might operate in parallel, perhaps in integral arrays developed using wafer-scale integration techniques.

A Solution?

"Three-and-a-half years ago, we concluded that it would be possible to build microscale gas turbines at very low cost using microfabrication techniques," said Epstein, who is a professor of aeronautics and astronautics as well as an ASME member. Following initial discussions, MIT's Lincoln Laboratory in Lexington, Mass., provided some seed money to start research on the topic, which Epstein admitted was "a solution looking for a problem at the time." The concept drew the interest of the U.S. Army Research Office in Research Triangle Park, N.C., which funded the group with $5 million to $6 million over five years (through 2000) to develop microturbines for mobile electric-power generation.

To date, the 25-member team has studied the basic fundamentals; developed a baseline design; and performed the necessary research on components, materials, and fabrication methods. The team is now conducting component testing, which should be complete in 1998, Epstein said. The final detailed design and initial engine fabrication will follow.

"It is by no means a foregone conclusion that the project will be successful," he added, "but after investing about 15 man-years of effort, we're pretty convinced it can be done. We are, however, still uncertain exactly how long it will take."

The microturbine project presents new challenges in the mechanical and electrical engineering disciplines of fluid dynamics, structural mechanics, bearings and rotor dynamics, combustion, and electric machinery design. "It's equally hard work no matter what the specialty," Epstein said. In addition, a series of tough trade-offs has to be made among issues related to fabrication difficulty, structural design, heat transfer, fluid mechanics, and electric performance.

Downscaling Concerns

These and other considerations are no different at microscale and macroscale, but the physics and mechanics influencing the design of the components do change with scale. Therefore, the optimal detailed designs can be quite different, according to the MIT team members. Examples of these scaling effects include the viscous forces in the fluid (which are larger at microscale), surface-area-to-volume ratios (also larger at microscale), chemical reaction times (invariant), the electric-field strength that can be realized (higher at microscale), and manufacturing constraints (limited mainly to two- dimensional planar geometries).

During the past two years, the group in the Gas Turbine Lab has performed a careful scaling study of high-speed, rotating turbomachinery, which has shown that suitably designed microdevices are remarkably promising. Interestingly, the researchers' intuition concerning the worst of the potential technical barriers were not always on the mark.

"We thought initially that scaling effects were going to cause problems relating to the viscosity of the air (airflow appears more viscous at microscales), and in operating such a tiny combustor," Epstein said. "However, it turned out that more-familiar turbine-engineering issues such as ensuring rotor-dynamics stability posed greater difficulties."

Exactly how big should a microengine be? The requirements for many power-production applications favor a larger engine size: 50 to 100 watts. Viscous effects in the fluid flow and combustor residence time requirements also favor a larger engine size. On the other hand, semiconductor manufacturing technology sets size limits; the upper size limit is set mainly by the etching-depth capability, which is currently a few hundred microns, while the lower limit is set by feature resolution and aspect ratio capabilities of the production processes.

A conventional (macroscopic) gas turbine, Epstein said, consists of a compressor, a combustion chamber, and a turbine that is driven by the combustion exhaust and powers the compressor. The residual heat content in the exhaust stream produces thrust. A large-scale gas turbine with a 1-meter-diameter air intake generates power on the order of 100 megawatts, he said. Scaled down to millimeter size, tens of watts could be produced, provided that the power density (power per unit of airflow) can be maintained. "We call it a microturbine because it will have one-millionth of the mass flow of a large 100-megawatt-size gas turbine," said Epstein, who added that others have called the diminutive engine a meso-scale device.

Keys to Miniaturization

In general, the success of the miniature heat engine requires the development of three microscale technologies: rotating machinery, combustors, and high-temperature material fabrication. Design calculations indicate that the device, to have sufficient power density, needs combustor exit temperatures of 1,000°C to 1,500°C; rotor peripheral speeds of 300 to 600 meters per second, and rotating structures centrifugally stressed to several hundred megapascals; low friction bearings; high geometric tolerances and tight clearances between rotating and static parts; and thermal isolation of the hot and cold sections.

According to Epstein, the key to achieving high power density in rotating machinery is high peripheral speed, because the power transferred to the air follows the square of the peripheral speed. Therefore, the microdevice has to turn at same peripheral speed as a large turbine. High speed implies high centrifugal stress, he added, because "stress also goes like the square of the rotational speed, so a high power density means a highly stressed structure."

In conventional practice, rotor speed is constrained to several hundred meters per second by the strength-to-density ratio of high-temperature metal alloys. However, "the strength of brittle materials is scale-dependent due to a lower probability of flaws and higher-quality surfaces," Epstein said. Therefore, defect-free microfabricated materials are quite strong. "Materials at microscale are much better than at macroscale," he said. "They're very much more forgiving. For example, with single-crystal silicon, we can design structures using mechanical engineering criteria that would be unfeasible using standard metals such as titanium or superalloys."

In addition, their density is only half that of superalloys. Thus, compared with macroscopic materials, these nonmetallic materials have a superior strength-to-density ratio, so they can be spun to high speeds without the risk of fracture.

As the high-temperature performance of silicon is limited by its creep life, a more refractory material is needed for the practical microturbine effort. Work is under way at the lab to design and fabricate silicon carbide and silicon/silicon carbide hybrid structures by chemical vapor deposition of relatively thick silicon carbide layers (10 to 200 microns) over silicon molds. Epstein called this modified processing technique a promising approach.

High rotating speeds also require low-friction bearings. Here the cubic scaling of volume—and mass—combined with the quadratic scaling of areas (the so-called cube-square law) mean that the surface-area-to-weight ratio of a rotor is large at small scales. This implies that miniature air bearings can support large loads.

Rotating Wheels

The baseline microturbine looks a lot like German turbine-engine inventor Hans von Ohain's first engine, Epstein said—a single-spool turbojet. Generally speaking, the engineers traded performance for simplicity in determining the baseline microturbine design, he noted.

The highly integrated design consists of a supersonic radial flow compressor and turbine connected by a hollow shaft (to limit heat conduction). The cylindrical device is 12 millimeters in diameter and 3 millimeters long. During operation, the wheels would spin at 2.5 million rpm.

Upon start-up, gaseous hydrogen fuel is injected at the compressor exit and mixes with air as it flows radially outward to the flame holders. The combustor discharges radially inward to the turbine whose exhaust turns 90 degrees to exit the engine nozzle. The baseline engine is expected to have a pressure ratio of 4:1 and an airflow of 0.15 grams per second. A thin-film electric starter/generator will be mounted on a shroud over the compressor blades, and will be cooled by compressor discharge air. Compressor discharge air will also cool the structure to isolate the compressor thermally from the combustor and turbine. The rotor is supported on air bearings.

The MIT research group is also developing a turbine/generator and motor-driven compressor. The turbine/generator includes a disk rotor with a radial inflow turbine wheel on one side and a generator rotor on the other. The rim of the disk serves as the journal air bearing, and the axial loads are supported by air-bearing thrust plates on either side of the disk center. The rotor is supported by the stator and end plates, with channels to distribute air to the bearings. Designed to produce 20 watts of electrical power with a compressed air inlet, the unit is primarily intended as a tool for developing the bearings and rotor dynamics, turbine aerodynamics, and generator electromechanics.

The motor-driven compressor has a similar geometric arrangement to that of the turbine, with the airfoils modified to those of a radial outflow compressor and the electrical machine configured as a motor. The design application in this case is the pressurization of 100-watt fuel cells.

As the engine is scaled down linearly, a turbine's power-to-weight ratio increases in the same fashion. Epstein said that the development of a small-scale combustor seemed challenging at first because chemical reaction times are invariant with size and the tiny volume of the device provides limited flow residence time. However, the time required to mix the fuel and air is a major component of the total combustion time for conventional turbines, and these mixing times do scale. The MIT team has successfully demonstrated a 2-millimeter-long combustion chamber that is 40 times the relative size of a conventional combustor.

The baseline generator design, meanwhile, is a 500-pole planar electric induction machine mounted on the shroud of the compressor rotor. Epstein said that the use of an electrostatic generator unit, as opposed to an electromagnetic one, was necessary because "at macroscales, an electromagnetic device has much greater power density. At microscales, we can use its electrostatic analog, since the electric-field strength is very much greater at short distances, so the power density of an electrostatic induction motor/generator can be similar." He described the microturbine generator as "a stepping-motor generator" with the rotor composed of a uniform conductor and the stator arranged in pie-shaped segments.

Microfabrication Methods

Researchers plan to demonstrate the manufacture of all the microengine components in silicon first, even as they develop the needed silicon carbide microfabrication methods. For the silicon components, Epstein said, "we're using established microfabrication techniques—masking, chemical vapor deposition, etching, the bonding of wafers into complicated structures, and so forth. For the silicon carbide parts, we'll still use quasi-standard microfabrication methods."

According to the MIT team, current micromachining technology primarily uses lithography to define planar geometries, which are then formed into prismatic structures by etching or vapor deposition. Multiple layers are used to create more-complex three-dimensional devices. The approach offers very high geometric (submicron) accuracies, parallel fabrication of large numbers of identical devices (so unit cost will be low), and simultaneous fabrication of both mechanical and electrical elements.

The MIT researchers face several challenges to successful microfabrication, including the production of features hundreds of microns deep, fillets to reduce stress on highly loaded parts, the correct electrical properties for the motor/generator, the excavation of volumes millimeters across and hundreds of microns deep, and assembly and packaging issues.

Modules of the MIT MEMCAD program, a process design tool for microelectromechanical systems, were used to create a "strawman" process for complete engines. Using only known process steps, the process yields wafers of completed engines, including a freely turning rotor, without additional assembly. It is a complex and aggressive process requiring seven aligned wafer bonds, 20 lithography steps, and the deposition of nine thin-film layers. The process complexity is roughly equivalent to doing a state-of-the-art complementary-metal-oxide-semiconductor integrated circuit.

As the first big step in the project, deep reactive ion etching was used to produce the 4-millimeter turbine wheel, which features blades with spans of 200 microns, and a 10-micron bearing gap between the rotor disk and the stator base plate. A test assembly also has been built at the die level, resulting in a test rig that includes thrust bearings, bearing air supply and interconnections, turbine inlet and exhaust ducting, and instrumentation provisions. Lab director Epstein noted that the silicon prototype will require external cooling to avoid premature failure.

The cube-square law yields another remarkable benefit that results from downsizing. If the power per unit of airflow remains constant (for the baseline engine, about 130 watts per gram per second), then the power-to-weight ratio would increase linearly as the size is reduced, because the airflow scales with the square of a linear dimension whereas the mass scales with the cube. If all else is kept the same as the engine is scaled down linearly, the airflow (and thus the power) decreases with the intake area (the square of the linear size), so that the power-to-weight ratio increases linearly as the engine size is reduced.

Detailed calculations indicate that the actual scaling is not quite that dramatic, but a millimeter-size engine would have a thrust-to-weight ratio of about 100:1, compared with 10:1 for the best modern aircraft engines.