Proceedings of ASME Turbo Expo 2003
Power for Land, Sea, and Air
June 16-19, 2003, Atlanta, Georgia, USA
GT-2003-38866
MILLIMETER-SCALE, MEMS GAS TURBINE ENGINES
Alan H. Epstein
Gas Turbine Laboratory
Massachusetts Institute of Technology
Cambridge, MA 02139 USA
[email protected]
ABSTRACT several of which are marketed commercially [1, 2]. Gas turbines
The confluence of market demand for greatly improved below a few hundred kilowatts in size generally use centrifugal
compact power sources for portable electronics with the rapidly turbomachinery (often derivative of automotive turbocharger
expanding capability of micromachining technology has made technology in the smaller sizes), but are otherwise very similar
feasible the development of gas turbines in the millimeter-size to their larger brethren in that they are fabricated in much the
range. With airfoil spans measured in 100’s of microns rather same way (cast, forged, machined, and assembled) from the
than meters, these “microengines” have about 1 millionth the same materials (steel, titanium, nickel superalloys). Recently,
air flow of large gas turbines and thus should produce about 1 manufacturing technologies developed by the semiconductor
millionth the power, 10-100 W. Based on semiconductor indus- industry have opened a new and very different design space for
try-derived processing of materials such as silicon and silicon gas turbine engines – one that enables gas turbines with diam-
carbide to submicron accuracy, such devices are known as eters of millimeters rather than meters, with airfoil dimensions
micro-electro-mechanical systems (MEMS). Current millime- in microns rather than millimeters. These shirt-button-sized gas
ter-scale designs use centrifugal turbomachinery with pressure turbine engines are the focus of this review.
ratios in the range of 2:1 to 4:1 and turbine inlet temperatures of Interest in millimeter-scale gas turbines is fueled by both
1200-1600 K. The projected performance of these engines are a technology push and a user pull. The technology push is the
on a par with gas turbines of the 1940’s. The thermodynamics of development of micromachining capability based on semicon-
MEMS gas turbines are the same as those for large engines but ductor manufacturing techniques. This enables the fabrication of
the mechanics differ due to scaling considerations and manufac- complex small parts and assemblies – devices with dimensions
turing constraints. The principal challenge is to arrive at a design in the 1-10,000 µm size range with submicron precision. Such
which meets the thermodynamic and component functional parts are produced with photolithographically-defined features
requirements while staying within the realm of realizable micro- and many can be made simultaneously, offering the promise of
machining technology. This paper reviews the state-of-the-art of low production cost in large-scale production. Such assemblies
millimeter-size gas turbine engines, including system design and are known in the US as micro-electrical-mechanical systems
integration, manufacturing, materials, component design, acces- (MEMS) and have been the subject of thousands of publica-
sories, applications, and economics. It discusses the underlying tions over the last two decades [3]. In Japan and Europe, devices
technical issues, reviews current design approaches, and dis- of this type are known as “microsystems”, a term which may
cusses future development and applications. encompass a wider variety of fabrication approaches. Early work
in MEMS focused on sensors and simple actuators, and many
INTRODUCTION devices based on this technology are in large-scale production,
For most of the 60-year-plus history of the gas turbine, such as pressure transducers and airbag accelerometers for auto-
economic forces have directed the industry toward ever larger mobiles. More recently, fluid handling is receiving attention.
engines, currently exceeding 100,000 lbs of thrust for aircraft For example, MEMS valves are commercially available, and
propulsion and 400 MW for electric power production applica- there are many emerging biomedical diagnostic applications.
tions. In the 1990’s, interest in smaller-size engines increased, Also, chemical engineers are pursing MEMS chemical reactors
in the few hundred pound thrust range for small aircraft and (chemical plants) on a chip [4].
missiles and in the 20-250 kW size for distributed power pro- User pull is predominantly one of electric power. The prolif-
duction (popularly known as “microturbines”). More recently, eration of small, portable electronics – computers, digital assis-
interest has developed in even smaller size machines, 1-10 kW, tants, cell phones, GPS receivers, etc. – require compact energy
1 Copyright ©2003 by ASME
supplies. Increasingly, these electronics demand energy supplies electric generator. A macroscale gas turbine with a meter-diame-
whose energy and power density exceed that of the best batteries ter air intake area generates power on the order of 100 MW. Thus,
available today. Also, the continuing advance in microelectron- tens of watts would be produced when such a device is scaled to
ics permits the shrinking of electronic subsystems of mobile millimeter size if the power per unit of air flow is maintained.
devices such as ground robots and air vehicles. These small, and When based on rotating machinery, such power density requires
in some cases very small, mobile systems require increasingly combustor exit temperatures of 1200-1600 K; rotor peripheral
compact power and propulsion. Hydrocarbon fuels burned in air speeds of 300-600 m/s and thus rotating structures centrifugally
have 20-30 times the energy density of the best current lithium stressed to several hundred MPa since the power density of both
chemistry-based batteries, so that fuelled systems need only be turbomachinery and electrical machines scale with the square of
modestly efficient to compete well with batteries. the speed, as does the rotor material centrifugal stress; low fric-
Given the need for high power density energy conversion in tion bearings; tight geometric tolerances and clearances between
very small packages, a millimeter-scale gas turbine is an obvi- rotating and static parts to inhibit fluid leakage, the clearances
ous candidate. The air flow through gas turbines of this size is in large engines are maintained at about one part in 2000 of the
about six orders of magnitude smaller than that of the largest diameter; and thermal isolation of the hot and cold sections.
engines and thus they should produce about a million times less These thermodynamic considerations are no different
power, 10-100 watts with equivalent cycles. Work first started on at micro- than at macroscale. But the physics and mechan-
MEMS approaches in the mid 1990’s [5-7]. Researchers rapidly ics influencing the design of the components do change with
discovered that gas turbines at these small sizes have no fewer scale, so that the optimal detailed designs can be quite different.
engineering challenges than do conventional machines and that Examples of these differences include the viscous forces in the
many of the solutions evolved over six decades of technology fluid (larger at microscale), usable strength of materials (larger at
development do not apply in the new design space. This paper microscale), surface area-to-volume ratios (larger at microscale),
reviews work on MEMS gas turbine engines for propulsion and chemical reaction times (invariant), realizable electric field
power production. It begins with a short discussion of scaling strength (higher at microscale), and manufacturing constraints
and preliminary design considerations, and then presents a con- (limited mainly to two-dimensional, planar geometries given
cise overview of relevant MEMS manufacturing techniques. In current microfabrication technology).
more depth, it examines the microscale implications for cycle There are many thermodynamic and architectural design
analysis, aerodynamic and structural design, materials, bearings choices in a device as complex as a gas turbine engine. These
and rotor dynamics, combustion, and controls and accessories. involve tradeoffs among fabrication difficulty, structural design,
The gas turbine engine as a system is then considered. This heat transfer, and fluid mechanics. Given a primary goal of
review then discusses propulsion and power applications and demonstrating that a high power density MEMS heat engine is
briefly looks at derivative technologies such as combined cycles, physically realizable, MIT’s research effort adopted the design
cogeneration, turbopumps, and rocket engines. The paper con- philosophy that the first engine should be as simple as possible,
cludes with thoughts on future developments. with performance traded for simplicity. For example, a recuper-
ated cycle, which requires the addition of a heat exchanger trans-
THERMODYNAMIC AND SCALING CONSIDERATIONS ferring heat from the turbine exhaust to the compressor discharge
Thermal power systems encompass a multitude of technical fluid, offers many benefits including reduced fuel consumption
disciplines. The architecture of the overall system is determined and relaxed turbomachinery performance requirements, but it
by thermodynamics while the design of the system’s components introduces additional design and fabrication complexity. Thus,
is influenced by fluid and structural mechanics and by material, the first designs are simple cycle gas turbines.
electrical and fabrication concerns. The physical constraints How big should a “micro” engine be? A micron, a milli-
on the design of the mechanical and electrical components are meter, a centimeter? Determination of the optimal size for such
often different at microscale than at more familiar sizes so that a device involves considerations of application requirements,
the optimal component and system designs are different as well. fluid mechanics and combustion, manufacturing constraints, and
Conceptually, any of the thermodynamic systems in use today economics. The requirements for many power production appli-
could be realized at microscale. Brayton (air) cycle and the Ran- cations favor a larger engine size, 50-100 W. Viscous effects
kine (vapor) cycle machines are steady flow devices while the in the fluid and combustor residence time requirements also
Otto [8], Diesel, and Stirling cycles are unsteady engines. The favor larger engine size. Current semiconductor manufacturing
Brayton power cycle (gas turbine) is superior based on consider- technology places both upper and lower limits on engine size.
ations of power density, simplicity of fabrication, ease of initial The upper size limit is set mainly by etching depth capability,
demonstration, ultimate efficiency, and thermal anisotropy. a few hundred microns at this time. The lower limit is set by
A conventional, macroscopic gas turbine generator consists feature resolution and aspect ratio. Economic concerns include
of a compressor, a combustion chamber, and a turbine driven by manufacturing yield and cost. A wafer of fixed size (say 200 mm
the combustion exhaust that powers the compressor. The residual diameter) would yield many more low power engines than high
enthalpy in the exhaust stream provides thrust or can power an power engines at essentially the same manufacturing cost per
2 Copyright ©2003 by ASME
1.4 important at small scale. Pressure ratios of 2:1 to 4:1 per stage
Specific Fuel Consumption (g/w/hr)
imply turbomachinery tip Mach numbers that are in the high
1.2 0.6 1200°K subsonic or supersonic range. Airfoil chords on the order of a
Co
m bus millimeter imply that a device with room temperature inflow,
t or
1.0 Exi
t Te
such as a compressor, will operate at Reynolds numbers in the
1300°K mp
era tens of thousands. With higher gas temperatures, turbines of
Co
tur
e similar size will operate at a Reynolds number of a few thou-
mp Ef
0.8 1400°K
res fici
0.7 1500°K sand. These are small values compared to the 105-106 range of
so enc
1600°K
r Is y
0.6 large-scale turbomachinery and viscous losses will be concomi-
en
0.8 tantly larger. But viscous losses make up only about a third of the
tro
pic
0.9
0.4
total fluid loss in a high speed turbomachine (3-D, tip leakage,
10 20 30 40 50 and shock wave losses account for most of the rest) so that the
Shaft Power Output (watts) per mm2 Inlet Area decrease in machine efficiency with size is not so dramatic. The
increased viscous forces also mean that fluid drag in small gaps
Figure 1: Simple cycle gas turbine performance with H2 fuel.
and on rotating disks will be relatively higher. Unless gas flow
passages are smaller than one micron, the fluid behavior can be
wafer. (Note that the sum of the power produced by all of the represented as continuum flow so that molecular kinetics, Knud-
engines on the wafer would remain constant at 1-10 kW.) When sen number considerations, are not important.
commercialized, applications and market forces may establish a Heat transfer is another aspect of fluid mechanics in which
strong preference here. For the first demonstrations of a concept, microdevices operate in a different design space than large-scale
a minimum technical risk approach is attractive. Analysis sug- machines. The fluid temperatures and velocities are the same but
gested that fluid mechanics would be difficult at smaller scales, the viscous forces are larger, so the fluid film heat transfer coef-
so the largest size near the edge of current microfabrication tech- ficients are higher by a factor of about 3. Not only is there more
nology, about a centimeter in diameter, was chosen as a focus of heat transfer to or from the structure but thermal conductance
MIT’s efforts. within the structure is higher due to the short length scale. Thus,
Performance calculations indicate that the power per unit air temperature gradients within the structure are reduced. This is
flow from the configuration discussed below is 50-150 W/(g/sec) helpful in reducing thermal stress but makes thermal isolation
of air flow (Figure 1). For a given rotor radius, the air flow rate challenging.
is limited primarily by airfoil span as set by stress in the turbine For structural mechanics, it is the change in material proper-
blade roots. Calculations suggest that it might be possible to ties with length scale that is most important. Very small length
improve the specific work, fuel consumption, and air flow rate scale influences both material properties and material selection.
in later designs with recuperators to realize microengines with In engines a few millimeters in diameter, design features such as
power outputs of as much as 50-100 W, power specific fuel blade tips, fillets, orifices, seals, etc may be only a few microns
consumption of 0.3-0.4 g/w-hr, and thrust-to-weight ratios of in size. Here, differences between mechanical design and mate-
100:1. This level of specific fuel consumption approaches that rial properties begin to blur. The scale is not so small (atomic
of current small gas turbine engines but the thrust-to-weight lattice or dislocation core size) that continuum mechanics no
ratio is 5-10 times better than that of the best aircraft engine. longer applies. Thus, elastic, plastic, heat conduction, creep,
The extremely high thrust-to-weight ratio is simply a result of and oxidation behaviors do not change, but fracture strength
the so-called “cube-square law”. All else being the same as can differ. Material selection is influenced both by mechanical
the engine is scaled down linearly, the air flow and thus the requirements and by fabrication constraints. For example, struc-
power decreases with the intake area (the square of the linear ture ceramics such as silicon carbide (SiC) and silicon nitride
size) while the weight decreases with the volume of the engine (Si3N4) have long been recognized as attractive candidates for
(the cube of the linear size), so that the power-to-weight ratio gas turbine components due to their high strength, low density,
increases linearly as the engine size is reduced. Detailed calcula- and good oxidation resistance. Their use has been limited, how-
tions show that the actual scaling is not quite this dramatic since ever, by the lack of technology to manufacture flaw-free material
the specific power is lower at the very small sizes [5]. A principal in sizes large enough for conventional engines. Shrinking engine
point is that a micro-heat engine is a different device than more size by three orders of magnitude virtually eliminates this prob-
familiar full-sized engines, with different weaknesses and differ- lem. Indeed, mass-produced, single-crystal semiconductor mate-
ent strengths. rials are essentially perfect down to the atomic level so that their
usable strength is an order of magnitude better than conventional
Mechanics Scaling metals. This higher strength can be used to realize lighter struc-
While the thermodynamics are invariant down to this scale, tures, higher rotation speeds (and thus higher power densities)
the mechanics are not. The fluid mechanics, for example, are at constant geometry, or simplified geometry (and thus manu-
scale-dependent [9]. One aspect is that viscous forces are more facturing) at constant peripheral speed. An additional material
3 Copyright ©2003 by ASME
105 Starting Compressor Diffuser
air in Thrust Inlet rotor vane Combustor
Critical Temperature Change (K)
h = 10,000 W/m2K bearing
3.7 mm
104 SiC
Nozzle guide vane
Journal Exhaust Turbine
Si3N4 bearing rotor
21 mm
103
Al2O3 Figure 3: H2 demo engine with conduction-cooled turbine
constructed from six silicon wafers.
102
10-3 10-2 10-1 The centrifugal compressor and radial turbine rotor diameters
Characteristic Length (m) are 8 mm and 6 mm respectively. The compressor discharge air
wraps around the outside of the combustor to cool the combustor
Figure 2: Critical temperature change to cause fracture via walls, capturing the waste heat and so increasing the combus-
thermal shock. tor efficiency while reducing the external package temperature.
The rotor radial loads are supported on a journal bearing on the
periphery of the compressor. Thrust bearings on the centerline
consideration is that thermal shock susceptibility decreases as
and a thrust balance piston behind the compressor disk support
part size shrinks. Thus, materials such as alumina (Al2O3) which
the axial loads. The balance piston is the air source for the hydro-
have very high temperature capabilities but are not considered
static journal bearing pressurization. The thrust bearings and bal-
high temperature structural ceramics due to their susceptibility to
ance piston are supplied from external air sources. The design
thermal shock are viable at millimeter length scales (Figure 2).
peripheral speed of the compressor is 500 m/s so that the rota-
Since these have not been considered as MEMS materials in the
tion rate is 1.2 Mrpm. External air is used to start the machine.
past, there is currently little suitable manufacturing technology
With 400 µm span airfoils, the unit is sized to pump about 0.36
available [10].
grams/sec of air, producing 0.1 Newtons of thrust or 17 watts of
shaft power. A cutaway engine chip is shown in Figure 4. In this
OVERVIEW OF A MEMS GAS TURBINE ENGINE DESIGN
particular engine build, the airfoil span is 225 µm and the disks
Efforts at MIT were initially directed at showing that a
are 300 μm thick.
MEMS-based gas turbine is indeed possible, by demonstrating
The following sections elaborate on the component tech-
benchtop operation of such a device. This implies that, for a
nologies of this engine design. It starts with a primer on micro-
first demonstration, it would be expedient to trade engine per-
fabrication and then goes on to turbomachinery aerodynamic
formance for simplicity, especially fabrication simplicity. Most
design, structures and materials, combustion, bearings and rotor
current, high precision, microfabrication technology applies
dynamics, and controls and accessories. A system integration
mainly to silicon. Since Si rapidly loses strength above 950 K,
discussion then expands on the high-level tradeoffs which define
this becomes an upper limit to the turbine rotor temperature.
the design space of a MEMS gas turbine engine.
But 950 K is too low a combustor exit temperature to close
the engine cycle (i.e. produce net power) with the component
A PRIMER ON MICROMACHINING
efficiencies available, so cooling is required for Si turbines. The
Gas turbine engine design has always been constrained by
simplest way to cool the turbine in a millimeter-sized machine
the practicality of manufacturing parts in the desired shape and
is to eliminate the shaft, and thus conduct the turbine heat to the
size and with the material properties needed. As with conven-
compressor, rejecting the heat to the compressor fluid. This has
the great advantage of simplicity and the great disadvantage of
lowering the pressure ratio of the now non-adiabatic compres-
sor from about 4:1 to 2:1 with a concomitant decrease in cycle
power output and efficiency. Hydrogen was chosen as the first
fuel to simplify the combustor development. This expedient
arrangement was referred to as the H2 demo engine. It is a gas
generator/turbojet designed with the objective of demonstrating
the concept of a MEMS gas turbine. It does not contain electrical
machinery or controls, all of which are external.
The MIT H2 demo engine design is shown in Figure 3. Figure 4: Cutaway H2 demo gas turbine chip.
4 Copyright ©2003 by ASME
tional metal fabrication, the mechanical and electrical properties wafer is carried out in parallel, leading to one of the great advan-
of MEMS materials can be strongly influenced by the fabrication tages of this micromachining approach, low unit cost. To greatly
process. simplify a complex process with very many options, the devices
While an old-school designer may have admonished his of interest will serve as illustrative examples.
team “Don’t let the manufacturing people tell you what you can’t First, the wafers are coated with a light-sensitive photore-
do!”, design for manufacturing is now an important concern in sist. A high contrast black-and-white pattern defining the geom-
industry. Major decisions in engine architecture are often set by etry is then optically transferred to the resist either by means of
manufacturing constraints. Of course this was true in the design a contact exposure with a glass plate containing the pattern (a
of Whittle’s first jet engine, in which the prominent external, “mask”), or by direct optical or e-beam writing. The photoresist
reverse flow combustors reflected the need to keep the turbine is then chemically developed as though it were photographic
very close to the compressor to control rotor dynamics given film, baked, and then the exposed areas are removed with a
that the forging technology of the day could only produce short, solvent. This leaves bare silicon in the areas to be etched and
small diameter shafts integral with a disk [11]. photoresist-protected silicon elsewhere. The etching process is
Compared to manufacturing technologies familiar at large based on the principle that the bare silicon is etched at a much
scale, current microfabrication technology is quite constrained higher rate, typically 50-100x, than the mask material. Many dif-
in the geometries that can be produced and this severely limits ferent options for making masks have been developed, including
engine design options. Indeed, the principal challenge is to arrive a wide variety of photoresists and various oxide or metal films.
at a design which meets the thermodynamic and component func- By using several layers of masking material, each sensitive to
tional requirements while staying within the realm of realizable different solvents, multi-depth structures can be defined. Photo-
micromachining technology. The following paragraphs pres- resist on top of SiO2 is one example.
ent a simple overview of current micromachining technology The exposed areas of the wafer can now be etched, either
important to this application and then discuss how it influences chemically or with a plasma. The devices we are concerned with
the design of very small rotating machinery. These technologies here require structures which are 100’s of microns high with very
were derived from the semiconductor industry 15-20 years ago, steep walls, thus a current technology of great interest is deep
but the business of micromachining has now progressed to the reactive ion etching (DRIE). In the DRIE machine, the wafer
level that considerable process equipment (known as “tools”) is is etched by plasma-assisted fluorine chemistry for several tens
developed specifically for these purposes [12]. of seconds, then the gas composition is changed and a micron
The primary fabrication processes important in this applica- or so of a teflon-like polymer is deposited which preferentially
tion are etching of photolithographically-defined planar geome- protects the vertical surfaces, and then the etch cycle is repeated
tries and bonding of multiple wafers. The usual starting point is a [13]. The average depth of a feature is a function of the etch time
flat wafer of the base material, most often single-crystal silicon. and the local geometry. The etch anisotropy (steepness of the
These wafers are typically 0.5 to 1.0 mm thick and 100 to 300 walls) can be changed by adjusting the plasma properties, gas
mm in diameter, the larger size representing the most modern composition, and pressure. In addition, these adjustments may
technology. Since a single device of interest here is typically alter the uniformity of the etch rate across the wafer by a few
a centimeter or two square, dozens to hundreds fit on a single percent since no machine is perfect. One feature of current tools
wafer (Figure 5). Ideally, the processing of all the devices on a is that the etch rate is a function of local geometry such as the
Figure 6: A 4:1 pressure ratio, 4 mm rotor dia radial inflow
Figure 5: Si wafer of radial inflow turbine stages. turbine stage.
5 Copyright ©2003 by ASME
lateral extent of a feature. This means that, for example, different current state-of-the-art, the airfoil length can be controlled to
width trenches etch at different rates, presenting a challenge to better than 1 µm across the disk, which is sufficient to achieve
the designer of a complex planar structure. A DRIE tool typi- high-speed operation without the need for dynamic balancing.
cally etches silicon at an average rate of 1-3 µm per minute, the Turbomachines of similar geometry have been produced with
precise rate being feature- and depth-dependent. Thus, structures blade spans of over 400 µm.
that are many hundreds of microns deep require many hours of The processing of a 4-mm-diameter turbine stage is illus-
etching. Such a tool currently costs $0.5-1.0M and etches one trated in Figure 7 as a somewhat simplified example. Note that
wafer at a time, so the etching operation is a dominant factor the vertical scaling in the figure is vastly exaggerated for clarity
in the cost of producing such deep mechanical structures. Both since the thickness of the layers varies so much (about 1 µm of
sides of a wafer may be etched sequentially. silicon dioxide and 10 µm of photoresist on 450 µm of silicon).
Figure 6 is an image of a 4 mm rotor diameter, radial inflow It is a 16-step process for wafer 1, requiring two photo masks.
air turbine designed to produce 60 watts of mechanical power It includes multiple steps of oxide growth (to protect the surface
at a tip speed of 500 m/s [14, 15]. The airfoil span is 200 µm. for wafer bonding), patterning, wet etching (with a buffered
The cylindrical structure in the center is a thrust pad for an axial hydrofluoric acid solution known as BOE), deep reactive ion
thrust air bearing. The circumferential gap between the rotor etching (DRIE), and wafer bonding (of the rotor wafer, #1, to an
and stator blades is a 15 µm wide air journal bearing required to adjoining wafer, #2, to prevent the rotor from falling out during
support the radial loads. The trailing edge of the rotor blades is processing). Note that wafer 2 in the figure was previously pro-
25 µm thick (uniform to within 0.5 µm) and the blade roots have cessed since it contains additional thrust bearing and plumbing
10 µm radius fillets for stress relief. While the airfoils appear features which are not shown here for clarity, In fact, it is more
planar in the figure, they are actually slightly tapered from hub complex to fabricate than the rotor wafer illustrated.
to tip. Current technology can yield a taper uniformity of about The second basic fabrication technology of interest here is
30:1 to 50:1 with either a positive or negative slope. At the the bonding together of processed wafers in precision alignment
UV Light
Wafer 1 Wafer 1
(a) 450 µm thick, 4 inch double-
side polished silicon wafer. (g) Wet oxide etch with liquid
Buffered Oxide Etch (BOE). Wafer 1
Wafer 1 Wafer 2
(b) 0.5 µm-thick-thermal oxidation. Wafer 1 (m) UV exposure photoresist.
(h) DRIE etch airfoils. Wafer 1
Wafer 1
Thrust Blades Vanes Wafer 2
bearing
(c) Spin-coat on ~10 µm-thick
(n) Develop photoresist.
photoresist. Wafer 1
UV Light (i) Remove photoresist. Wafer 1
Glass Wafer 2
mask Wafer 1
Wafer 1 (o) Oxide patterning by BOE
Wafer 2
(d) UV exposure photoresist.
(j) Remove oxide on bonding side. Wafer 1
Wafer 2
Wafer 1 Wafer 1
Wafer 2 (p) DRIE etch of journal bearing.
(e) Develop photoresist.
(k) Direct silicon bond 1 to 2. Journal bearing
Wafer 1
Wafer 1 Wafer 1 Wafer 2
(f) Protect back-side oxide Wafer 2 (q) Strip photoresist and oxide.
with photoreist. (l) Spin-coat on ~20 µm-thick photoresist. Ready for full-stack bonding.
Si Oxide Photoresist (Courtesy of N. Miki)
Figure 7: Simplified processing steps to produce the turbine in Figure 6 in a wafer stack.
6 Copyright ©2003 by ASME
so as to form multilayer structures. There are several classes of arrange a sequence of fabrication steps with all processing done
wafer bonding technologies. One uses an intermediate bonding at the wafer level so that a freely-rotating captured rotor is the
layer such as a gold eutectic or SiO2. These approaches, however, end product. The process must be such that the rotor is not free at
result in structures which have limited temperature capabilities, any time during which it can fall out, i.e. it must be mechanically
a few hundred °C. It is also possible to directly bond silicon to constrained at all times. There are several ways that this can be
silicon and realize the material’s intrinsic strength through the accomplished. For example, the layer containing the rotor can
entire usable temperature range of the material [16, 17]. Direct be “glued” to adjoining wafers with an oxide during fabrication.
bonding requires very smooth (better than 10 nanometers) and This glue can then be dissolved away to free the rotor after the
very clean surfaces (a single 1-µm-diameter particle can keep device is completely fabricated. In one version of the 4 mm tur-
several square centimeters of surface from bonding). Thus, a bine of Figure 6, an SiO2 film bonds the rotor layer at the thrust
very high standard of cleanliness and wafer handling must be bearing pad to the adjoining wafer, before the journal bearing is
maintained throughout the fabrication process. The wafers to etched. Another approach employs “break-off tabs” or mechani-
be bonded are hydrated and then aligned using reference marks cal fuses, flimsy structures which retain the rotor in place during
previously etched in the surface. The aligned wafers are brought fabrication and are mechanically failed after fabrication is com-
into contact and held there by Van der Waals forces. The stack of plete to release the rotor [19]. Both approaches have been proven
wafers is then pressed and heated to a few hundred degrees for successful.
tens of minutes. Finally, the stack is annealed for about one hour The last MEMS technology we will mention is that for
at 1100°C in an inert gas furnace. (If a lower temperature is used, electronic circuitry, mainly for embedded sensors and elec-
a much longer time will be needed for annealing.) Such a stack, tric machinery such as actuators, motors, and generators. The
well-processed from clean wafers, will not have any discernable circuitry is generally constructed by laying down alternating
bond lines, even under high magnification. Tests show the bonds insulating and conducting layers, typically by using vapor depo-
to be as strong as the base material. The current state-of-the-art is sition or sputtering approaches, and patterning them as they are
stacks of 5-6 wafers aligned across a wafer to 0.5-1.0 µm. More applied using the photoresist technology outlined above. While
wafers can be bonded if alignment precision is less important. the microelectronics industry has developed an extremely wide
Note that the annealing temperature is generally higher than set of such technologies, only a small subset are compatible with
devices encounter in operation. This process step thus repre- the relatively harsh environment of the processing needed to
sents the limiting temperature for the selection of materials to realize wafer-bonded mechanical structures hundreds of microns
be included within the device [18]. Figure 8 shows the turbine deep. Specifically, the high wafer annealing temperatures limit
layer of Figure 6 bonded as the center of a stack of five wafers, the conductor choices to polysilicon or high temperature metals
the others contain the thrust bearings and fluid plumbing. such as platinum or tungsten. The energetic etching processes
A third fabrication technology of interest for micro-rotating require relatively thick masking material which limits the small-
machinery is that which realizes a freely-spinning rotor within a est electrical feature size to the order of a micron, rather than
wafer-bonded structure. We require completed micromachines the tens of nanometers used in state-of-the-art microelectronic
which include freely-rotating assemblies with clearances mea- devices.
sured in microns. While it is possible to separately fabricate Using the above technologies, shapes are restricted to mainly
rotors, insert them into a stationary structure, and then bond
Thrust-bearing
an overlaying static structure, this implies pick-and-place hand Rotor supply plenum Forward
operations (rather than parallel processing of complete wafers) Exhaust thrust bearing
and increases the difficulty in maintaining surfaces sufficiently
clean for bonding. A fundamentally different approach is to
Turbine Hydrostatic Journal Turbine
Stator Rotor Thrust Bearings Bearing blade
Journal
5 bearing
Rotor wafer
Static stack
Structure Side force
plenum
Journal Thrust Aft Journal Aft thrust
Pressurization Balance Exhaust Pressurization bearing
Plenum Plenum Plenum
(a) Conceptual Cross-Section (b) Electron Microscope Image of Cross-Section
Figure 8: Complete, 5-layer turbine “stack” including bearings and fluid plumbing.
7 Copyright ©2003 by ASME
prismatic or “extruded” geometries of constant height. Ongoing deleterious to the fluid flow. For example, at the 2-mm-diameter
research with greyscale lithography suggests that smoothly inlet to a compressor impeller, 3-D CFD simulations show that
variable etch depths (and thus airfoils of variable span) may be a right-angle turn costs 5% in compressor efficiency and 15%
feasible in the near term [20]. Conceptually, more complex 3- in mass flow compared to a smooth turn [21]. Engineering
D shapes can be constructed of multiple precision-aligned 2-D approaches to this problem include lowering the Mach number
layers. But layering is expensive with current technology and at the turns (by increasing the flow area), smoothing the turns
5-6 is considered a large number of precision-aligned layers for a with steps or angles (which adds fabrication complexity), and
microdevice. Since 3-D rotating machine geometries are difficult adding externally-produced contoured parts when the turns are
to realize, planar geometries are preferred. While 3-D shapes are at the inlet or outlet to the chip.
difficult, in-plane 2-D geometric complexity is essentially free
in manufacture since photolithography and etching process an Compressor Aerodynamic Design
entire wafer at one time. These are much different manufacturing The engine cycle demands pressure ratios of 2-4, the higher
constraints than are common in the large-scale world so it is not the better. This implies that transonic tip Mach numbers and
surprising the optimal machine design may also be different. therefore rotor tip peripheral speeds in the 400-500 m/s range
are needed. This yields Reynolds numbers (Re) in the range of
TURBOMACHINERY FLUID MECHANICS 104 for millimeter-chord blades. The sensitivity of 2-D blade
The turbomachine designs considered to date for MEMS performance to Re in this regime is illustrated in Figure 9, which
engine applications have all been centrifugal since this geometry presents the variations of efficiency with size for a radial flow
is readily compatible with manufacturing techniques involving compressor and turbine. While this analysis suggests that for low
planar lithography. (It is also possible to manufacture single loss it is desirable to maximize chord, note that the span of the
axial flow stages by using intrinsic stresses generated in the airfoils is less than the chord, implying that aero designs should
manufacturing process to warp what otherwise would be planar include endwall considerations at this scale.
paddles into twisted blades, but such techniques have not been In conventional size machines the flow path contracts to
pursued for high-speed turbomachinery). In most ways, the control diffusion. Since this was not possible with established
fluid mechanics of microturbomachinery are similar to that of micromachining technology, the first approach taken was to con-
large-scale machines. For example, high tip speeds are needed to trol diffusion in blade and vane passages by tailoring the airfoil
achieve high pressure ratios per stage. Micromachines are differ- thickness rather than the passage height [21, 22]. This approach
ent in two significant ways: small Reynolds numbers (increased results in very thick blades, as can be seen in the 4:1 pressure
viscous forces in the fluid) and, currently, 2-D, prismatic geom- ratio compressor shown in Figure 10. Compared to conventional
etry limitations. The low Reynolds numbers, 103-105, are simply blading, the trailing edges are relatively thick and the exit angle
a reflection of the small size, and place the designs in the laminar is quite high. The design trade is between thick trailing edges
or transitional range. These values are low enough that it is dif- (which add loss to the rotor) or high rotor exit angles (which
ficult to diffuse the flow, either in a rotor or a stator, without result in reduced work at constant wheel speed, increased dif-
separation. This implies that either most of the stage work must fuser loss, and reduced operating range).
come from the centrifugal pressure change or that some separa- Although the geometry is 2-D, the fluid flow is not. The
tion must be tolerated. The design challenges introduced by the relatively short blade spans, thick airfoil tips, and low Reynolds
low Reynolds numbers are exacerbated by geometric restrictions numbers result in large hub-to-tip flow variations, especially at
imposed by current microfabrication technology. In particular,
the fabrication constraint of constant passage height is a problem 6
Normalized Total Pressure Loss
in these high-speed designs. High work on the fluid means large Compressor
fluid density changes. In conventional centrifugal turboma- Turbine
5
chinery, density change is accommodated in compressors by Compressor
contracting or in turbines by expanding the height of the flow 4 Design Point
D
path. However, conventional microfabrication technology is not iff
us
amenable to tapering passage heights, so all devices built to date 3 er
Im
pe No
have a constant span. How these design challenges manifest lle zz
2 r le
themselves are somewhat different in compressors and turbines. Ro
tor
A common fluid design challenge is turning the flow to 1
angles orthogonal to the lithographically-defined etch plane, Turbine
such as at the impeller eye or the outer periphery of the compres- Design Point
0
sor diffuser. At conventional scale, these geometries would be 103 104 105 106
carefully contoured and perhaps turning vanes would be added. Reynolds Number
Such geometry is currently difficult to produce with microfabri- Figure 9: Calculated sensitivity of 2-D airfoil loss with
cation, which most naturally produces sharp right angles that are Reynolds number [9].
8 Copyright ©2003 by ASME
for designs in which the blade tip is at least as wide as the pas-
sage. The full-scale blading dimensions of the microcompressor
tested scaled-up was a blade chord of about 1000 µm and a span
of 225 µm. Thus the design minimum tip clearance of 2 µm (set
to avoid blade tip rubs) represents 0.2% of chord and 1% of span.
Figure 11 includes measurements of the sensitivity of this design
to tip clearance.
Recent microfabrication advances using greyscale lithog-
raphy approaches suggest that variable span turbomachinery
may indeed be feasible [20]. This would facilitate designs
with attached flow on thin blades. Compared to the thick blade
approach, 3-D CFD simulations of thin blade compressors with
a tip shroud show about twice the mass flow for the same maxi-
mum span and wheel speed, an increase in pressure ratio from
2.5 to 3.5, and an increase in adiabatic efficiency from 50% to
Figure 10: A 500 m/s tip speed, 8 mm dia centrifugal engine
70% [25].
compressor.
Isomura et al. have taken a different approach to centime-
ter-scale centrifugal compressors [26, 27]. They have chosen to
the impeller exit. This imposes a spanwise variation on stator scale a conventional 3-D aerodynamic machine with an inducer
inlet angle of 15-20 degrees for the geometries examined. This down to a 12 mm diameter for a design 2 g/s mass flow rate and
cannot be accommodated by twisting the airfoils, which is not 3:1 pressure ratio. The test article is machined from aluminum
permitted in current microfabrication. The limited ability to on a high-precision, five-axis miller. No test results have been
diffuse the flow without separation at these Reynolds numbers reported to date.
also precludes the use of vaneless diffusers if high efficiency is Kang et al. [28] have built a 12-mm-diameter conventional
required, since the flow rapidly separates off parallel endwalls. geometry centrifugal compressor from silicon nitride using a
While extensive 2-D and 3-D numerical simulations have rapid prototype technology known as mold shape deposition
been used to help in the design and analysis of the microma- manufacture. It was designed to produce a pressure ratio of 3:1
chines, as in all high-speed turbomachinery development, test at 500 m/s tip speed with a mass flow of 2.5 g/s and an efficiency
data is needed. Instrumentation suitable for fluid flow measure- of 65-70%. To date, they report testing up to 250 m/s and perfor-
ments in turbomachinery with blade spans of a few hundred mance consistent with CFD analysis.
microns and turning at over a million rpm is not readily avail- A major aerodynamic design issue peculiar to these very
able. While it is theoretically possible to microfabricate the small machines is their sensitivity to heat addition. It is difficult
required instrumentation into the turbomachine, this approach to design a centimeter-scale gas turbine engine to be completely
to instrumentation is at least as difficult as fabricating the micro-
turbomachine in the first place. Instead, the standard technique
of using a scaled turbomachine test rig was adopted [23]. In this 2.2
}
0.8% 2% 1%
case the test rig was a 75x linear scale-up of a 4-mm-diameter 2.5% Data 100%
compressor (sufficiently large with a 300-mm-rotor diameter for 2.0
6.7%
5%
conventional instrumentation) rather than the 2-4x scale-down 3-D CFD
Corrected Pressure Ratio
common in industry. The geometry tested was a model of a
2:1 pressure ratio, 4-mm-diameter compressor with a design tip 1.8
speed of 400 m/sec for use in a micromotor-driven air compres- 80%
sor [24]. This design used the thick-blade-to-control-diffusion 1.6
philosophy discussed above. The rig was operated at reduced
inlet pressure to match the full-scale design Reynolds number 1.4
of about 20,000. A comparison of a steady, 3-D, viscous CFD 60%
(FLUENT) simulation to data is shown in Figure 11. The simula-
tion domain included the blade tip gaps and right-angle turn at 1.2
the inlet. It predicts the pressure rise and mass flow rate to 5%
and 10%, respectively. 1.0
Tight clearances are considered highly desirable for com- 0 0.2 0.4 0.6 0.8 1.0 1.2
pressor aerodynamics in general but are a two-edged sword for Corrected Mass Flow (fraction of design)
the thick-bladed designs discussed above. Small tip clearance
Figure 11: Sensitivity of compressor pressure rise to tip
reduces leakage flow and its associated losses, but increases drag clearance (% span).
9 Copyright ©2003 by ASME
1.0
Efficiency (η/ηad) and Mass Flow (m/mad)
1.0 1.0
⋅ ⋅
0.9
Polytropic Efficiency
0.9 0.9
Pressure Ratio (π / πad)
πad 0.8
0.8 2 0.8
0.7
0.7 3 0.7
4 0.6
Engine data
0.6 0.6
5 0.5 Part-speed rig data
3-D CFD
0.5 0.5
10 0.4
10-4 10-3 10-2 10-1 100 101 102 103
0.4 0.4
0 0.2 0.4 0.6 0.8 1.0 1.2 Mass flow (Kg/sec)
Heat Transfer / Shaft Work (Q/W)
Figure 13: Variation of engine compressor polytropic
Figure 12: The influence of heat addition on compressor efficiency with size.
performance (pressure ratio is π, the subscript “ad” refers
to the adiabatic condition). in Figure 14. With a 400 µm span it is designed to produce 53 W
of shaft power at a pressure ratio (T-S) of 2.1, tip speed of 370
m/s, and mass flow of 0.28 g/s. The reaction is 0.2 which means
adiabatic, thus there will be some degree of heat addition through that the flow is accelerating through the turbine. Three-dimen-
conduction. An isothermal compressor at fixed temperature sional CFD simulations were used to explore the performance of
exhibits behavior close to that of an adiabatic machine with the this design using FLUENT. The calculational domain included
same amount of heat added at the inlet [29]. Thus, the influence the blade tip gap regions, the discharge of bearing air into the tur-
of the heat addition shows up as reductions in mass flow, pres- bine, and the right-angle turn and duct downstream of the rotor.
sure rise, and adiabatic efficiency. The effect of heat addition on These calculations predict that this design has an adiabatic effi-
compressor efficiency and pressure ratio are shown in Figure 12. ciency of about 60%. The remainder of the power goes to NGV
These effects can be quite dramatic at high levels of heat flow. losses (9%), rotor losses (11%), and exit losses (20%) [30]. These
The influence of this nonadiabatic behavior on the overall cycle are very low aspect ratio airfoils (~0.25) and this is reflected in
performance will be discussed later. the shear on the endwalls being about twice that on the airfoil
The ultimate efficiency potential for compressors in this size surfaces. The exit losses, the largest source of inefficiency, con-
range has yet to be determined. Figure 13 plots the polytropic sist of residual swirl, losses in the right-angle turn, and lack of
efficiency of a number of aeroengines and ground-based gas pressure recovery in the downstream duct. This implies that (a)
turbine compressors using inlet-corrected mass flow as an indi-
cator of size. The efficiency decreases with size but how much
of this is intrinsic to the fluid physics and how much is due to
the discrepancy in development effort (little engines have little
budgets) is not clear. (Note that there is an inconsistency of about
a percent in this data due to different definitions of efficiency,
i.e. whether losses in the inlet guide vanes and the exit vanes or
struts are included.)
Turbine Aerodynamic and Heat Transfer Design
While the aerodynamic design of a microfabricated, centi-
meter-diameter radial inflow turbine shares many of the design
challenges of a similar scale compressor, such as a constant
airfoil span manufacturing constraint, the emphasis is different.
Diffusion within the blade passages is not the dominant issue it is
with the compressor, so the thick blade shapes are not attractive.
The Reynolds numbers are lower, however, given increased vis-
cosity of the high temperature combustor exit fluid. The nozzle
guide vanes (NGVs) operate at a Re of 1,000-2,000 for millime- Figure 14: Silicon engine radial inflow turbine inside
ter-chord airfoils. annular combustor; the flow passages in the NGV’s are for
One 6-mm-diameter, constant-span engine turbine is shown bearing and balance air.
10 Copyright ©2003 by ASME
the rotor exit Mach number should be reduced if possible, and Table 1: A comparison of a microengine combustor with a
(b) that the turbine would benefit from an exit diffuser. large aeroengine combustor
High engine-specific powers require turbine inlet tempera-
Conventional Micro-
tures (TIT) above the 950 K capability of uncooled single-crystal
Combustor Combustor
Si. The MIT demo engine was designed with a TIT of 1600 K
and so requires turbine cooling. In the demo design the turbine Length 0.2 m 0.001 m
is conductively cooled through the structure. The heat flow is Volume 0.073 m3 6.6x10-8 m3
on the same order as the shaft power, and the resultant entropy Cross-sectional area 0.36 m2 6x10-5 m2
reduction is equivalent to 1-2% improvement in turbine effi- Inlet total pressure 37.5 atm 4 atm
ciency. Advanced engine designs may use film cooling. A major Inlet total temperature 870 K 500 K
issue in this case is the stability of a cold boundary layer on a Mass flow rate 140 kg/s 1.8x10-4 kg/s
rotating disk with radial inflow. While this is, in general, an Residence time ~7 ms ~0.5 ms
unstable flow, Philippon has shown through analysis and CFD Efficiency >99% >90%
simulation that the region of interest for these millimeter-scale Pressure ratio >0.95 >0.95
turbines lies in a stable regime (e.g. the boundary layers should Exit temperature 1800 K 1600 K
stay attached) [30]. He then designed film-cooled turbines and Power Density 1960 MW/m3 3000 MW/m3
analyzed these designs with CFD simulation. (Note: residence times are calculated using inlet pressure and
Based upon the work to date, it should be possible to realize an average flow temperature of 1000 K.)
microfabricated single-stage compressors with adiabatic pres-
sure ratios above 4:1 at 500 m/s tip speed with total-to-static
efficiencies of 50-60%. Achievable turbine efficiencies may be reduce the combustor efficiency and lower the reaction tem-
5-10% higher. perature. This narrows the flammability limits and decreases the
kinetic rates, which drops the effective Damkohler number. As
COMBUSTION an example, Figure 15 [31] illustrates the viable design space for
The primary design requirements for gas turbine combustors an H2-fuelled, 0.07 cc microcombustor as a function of the heat
include large temperature rise, high efficiency, low pressure drop, lost to the walls and as constrained by flame stability, structure
structural integrity, ignition, stability, and low emissions. These limits, and cycle requirement considerations. The design space
requirements are no different for a microcombustor which may shown permits a trade between heat loss and stoichiometry,
flow less than 1 g/s of air than for a 100 kg/sec large machine, which is especially important when burning hydrocarbons with
but the implementation required to achieve them can be. A com- narrow stoichiometry bounds.
parison between a modern aircraft engine combustor and a micro- The design details are dependent on the fuel chosen. The
engine is shown in Table 1 [31]. Scaling considerations result in design approach first taken was to separate the fuel-air mixing
the power density of a microcombustor exceeding that of a large from the chemical reaction. This is accomplished by premixing
engine. However, the combustor volume relative to the rest of the the fuel with the compressor discharge air upstream of the com-
microengine is much larger, by a factor of 40, than that of a large bustor flame holders. This permits a reduction of the combustor
engine. The reasons for this scaling can be understood in refer- residence time required by a factor of about 10 from the usual
ence to the basics of combustion science [32]. 5-10 msec. The disadvantage of this approach is a susceptibility
Combustion requires the mixing of fuel and air followed to flashback from the combustor into the premix zone, which
by chemical reaction. The time required to complete these
processes is generally referred to as the required combustion 0
residence time and effectively sets the minimum volume of the M
in
M
ax
-0.1 .C .T Th
combustor for a given mass flow. The mixing time can scale with yc u erm
Fl le DE rbi al S
device size but the chemical reaction times do not. In a large -0.2 am Te SI ne
G Te
tre
ss
e S mp N
engine, mixing may account for more than 90% of combustor -0.3
ta
bi , T SP mp
lit IT= AC , TI
Burn Lean
residence time. A useful metric is the homogeneous Damkohler y 16 E T=
18
Q (kW)
-0.4 00
number, which is the ratio of the actual fluid residence time in K 00
K
the combustor to the reaction time. Obviously a Damkohler of -0.5
Flammability
one or greater is needed for complete combustion and therefore
-0.6
high combustion efficiency. One difference between large and
microscale machines is the increased surface area-to-volume -0.7
ratio at small sizes. This offers more area for catalysts; it also -0.8
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
implies that microcombustors have proportionately larger heat
Equivalence Ratio ( )
losses. While combustor heat loss is negligible for large-scale
engines, it is a dominant design factor at microscale since it can Figure 15: Design space for Si H2 microcombustor.
11 Copyright ©2003 by ASME
must be avoided. To expedite the demonstration of a micro-gas in which dilution holes have been added to the liner creating a
turbine engine, hydrogen was chosen as the initial fuel because dual-zone combustor [31]. The missing data is due to instrumen-
of its wide flammability limits and fast reaction time. This is the tation burnout. The dual-zone configuration, in which the dilution
same approach taken by von Ohain when developing the first jets set up recirculation zones within the combustor, extends the
jet engine in Germany in the 1930’s. Hydrogen is particularly operating range by about a factor of two at a cost of 10-20% in
attractive because it will burn at equivalence ratios, φ, as low combustor efficiency. These combustors have been operated at
as 0.3 which yields adiabatic combustion temperatures below exit temperatures above 1800 K.
1500 K, facilitating the realization of simple premixed designs. Hydrocarbon fuels such as methane and propane have reac-
Microcombustor technology has been developed in several tion rates only about 20% of those of H2, requiring larger combus-
full-sized (i.e. micro) test rigs which duplicate the geometry of tor volumes for the same heat release. They also must react closer
an engine but with the rotating parts replaced with stationary to stoichiometric and therefore at higher temperatures, above
swirl vanes [33]. In the Si micromachined geometry of Figures 2000 K. For gas phase (homogeneous) combustion designs this
3 and 4, to reduce heat losses through the walls and therefore to requires a multizone burner (stoichiometric zone followed by a
increase combustor efficiency, the inlet air wraps around the out- dilution region) as used on most large gas turbines. Alternatively,
side of the 0.2 cc combustor before entering through flame hold- heterogeneous reactions on the surface of a catalyst can widen the
ers in a reversed flow configuration. This configuration is similar flammability limits and so reduce the combustion temperature.
to the traditional reverse-flow engine combustor but scaled down Both approaches have been demonstrated at microscale. Ethylene
to 0.1-0.3 g/sec air flow rate. The Si liner in this case is conduc- (which has a high reaction rate) and propane have been burned in
tion- rather than film-cooled. In this premixed approach, fuel is the H2 combustors described above. The combustion efficiency
injected near the inlet of the upstream duct to allow time for fuel- with ethylene approached 90% while that for propane was closer
air mixing without requiring additional combustor volume. This to 60%. These fuels need larger combustor volumes compared to
design takes advantage of microfabrication’s ability to produce hydrogen for the same heat release. Data for a variety of geome-
similar geometric features simultaneously, using 90 fuel injec- tries and fuels is reduced in terms of Damkohler number in Figure
tion ports, each 120 µm in diameter, to promote uniform fuel-air 17, which shows that values of greater than 2 are needed for high
mixing. A simple hot wire loop provides ignition [34]. chemical efficiency [31].
The combustor was tested in several configurations including Catalytic microcombustors have been produced by filling
variations of flame holder and dilution hole geometry. Combus- the combustor volume of the above geometries with a platinum-
tion efficiencies approaching 100% have been reported with pres- coated nickel foam. For propane, the catalyst increased the heat
sure ratios of about 0.95-0.98. The H2 data in Figure 16 shows the release in the same volume by a factor of 4-5 compared to the
variation of combustor efficiency versus mass flow rate for two propane-air results discussed above. Pressure drops through the
configurations, one purely premixed (no dilution holes) and one foam are only 1-2% [35]. Presumably catalytic combustor per-
formance can be improved by a better choice of catalyst (plati-
120 num was selected for H2) and a geometry optimized for catalytic
Dual-Zone
= 0.4 rather than gas-phase combustion.
= 0.6 Takahashi et al. [36] are developing combustors designed
Combustor
100 95%
confidence
interval
Combustor Efficiency
80 * * ** * *
1.0 * * ** *
* ** * *
* * * * *
* * *** * * *
Chemical efficiency
60 0.8 * ** * * *
** * * * *
** **
*
0.6 * *
40 **
Premix *
Combustor 0.4 * *
20 **
* Six-wafer (annular)
95% confidence Six-wafer (slotted)
0.2 *
* Dual-zone
Three-stack
0
0 0.05 0.1 0.15 0.2 0
0 2 4 6 8 10 12 14
Air Flow (g/sec) Damkohler No.
Figure 16: Measured performance of 0.2 cc, Si Figure 17: Measured microcombustor performance as a
microcombustors using H2 fuel. function of Damkohler number.
12 Copyright ©2003 by ASME
for somewhat larger gas turbines, with flow rates of about 2 g/s. able rolling contact bearings. A microfabricated bearing solution
Designed for methane, these are a miniature version of can-type is needed. Both electromagnetic and air bearings have been con-
industrial combustion chambers with a convection-cooled liner sidered for this application.
and dilution holes. These are conventionally machined with Electromagnetic bearings can be implemented with either
volumes of 2-4 cc. The combustion efficiencies of these units magnetic or electric fields providing the rotor support force.
have been demonstrated as above 99% at equivalence ratios of Although extensive work has been done on the application
about 0.37 with a design combustor exit temperature of 1323 K. of magnetic bearings to large rotating machinery, work is just
The design residence time is about 6.5 ms. Matsuo et al. [37] beginning on magnetic bearings for micromachines. In addi-
constructed a still larger (20 cc volume, 16 g/s flow rate) con- tion to their complexity, magnetic bearings have two major
ventionally-machined combustor burning liquefied natural gas. challenges in this application. First, magnetic materials are not
They report a combustor exit temperature of about 1200 K. compatible with most microfabrication technologies, limiting
Overall, experiments and calculations to date indicate device fabrication options. Second, Curie point considerations
that high efficiency combustion systems can be engineered limit the temperatures at which magnetic designs can operate to
at microscale and achieve the heat release rate and efficiency below those encountered in the micro-gas turbine, so consider-
needed for very small gas turbine engines. able cooling may be needed. For these reasons, the first efforts
concentrated on designs employing electric fields. The designs
BEARINGS AND ROTOR DYNAMICS examined did not appear promising in that the forces produced
The mechanical design of gas turbine engines is dominated were marginal compared to the bearing loads expected [38].
by the bearings and rotor dynamics considerations of high- Also, since electromagnetic bearings are unstable, feedback
speed rotating machinery. Micromachines are no different in stabilization is needed, adding to system complexity.
this regard. As in all high-speed rotating machinery, the basic Gas bearings support their load on thin layers of pressur-
mechanical architecture of the device must be laid out so as ized gas. For micromachines such as turbines they have intrinsic
to avoid rotor dynamic problems. The high peripheral speeds advantages over electromagnetic approaches, including no
required by the fluid and electromechanics lead to designs which temperature limits, high load bearing capability, and relative
are supercritical (operate above the natural resonant frequency of manufacturing simplicity. At large scale, gas bearings are used
the rotor system), just as they often are in large gas turbines. in many high-speed turbomachinery applications, including
Key design requirements imposed by the rotor dynamics aircraft environmental control units, auxiliary power units,
are that mechanical critical (resonant) frequencies lie outside 30-70 kW “microturbines”, and turbochargers [39]. At smaller
the steady-state operating envelope, and that any critical fre- scale, gas bearings have been used in gyroscopic instruments for
quencies that must be traversed during acceleration are of suf- many years. All else being the same, the relative load-bearing
ficiently low amplitude to avoid rubs or unacceptable vibrations. capability of a gas bearing improves as size decreases since the
The bearings play an important role in the rotor dynamics since volume-to-surface area ratio (and thus the inertial load) scales
their location and dynamical properties (stiffness and damping) inversely with size. Rotor and bearing dynamics scaling is more
are a major determinant of the rotor dynamics. The bearings in complex [40]. However, rotor dynamics in this application are
turn must support the rotor against all radial and axial loads seen somewhat simplified compared to large engines since the struc-
in service. In addition to the rotor dynamic forces, the bearing ture is very stiff, so only rigid body modes need be considered.
loads under normal operation include all the net pressure and In the following paragraphs we will first discuss journal bearings
electrical forces acting on the rotor as well as the weight of the which support radial loads and then consider thrust bearings
rotor times the external accelerations imposed on the device. For needed for axial loads.
aircraft engines this is usually chosen as 9 g’s, but a small device The simplest journal bearing is a cylindrical rotor within a
dropped on a hard floor from two meters experiences consider- close-fitting circular journal. Other, more complex, variations
ably larger peak accelerations. An additional requirement for used in large size machines include foil bearings and wave bear-
portable equipment is that the rotor support be independent of ings. These can offer several advantages but are more difficult to
device orientation. The bearing technology chosen must be com- manufacture at very small size. Thus, the plane cylindrical geom-
patible with the high temperatures in a gas turbine engine (or be etry was the first approach adopted since it seemed the easiest
protected within cooled compartments) and be compatible with to microfabricate. Gas bearings of this type can be categorized
the fabrication processes. into two general classes which have differing load capacities
Early MEMS rotating machines have been mainly micro- and dynamical characteristics. When the gas pressure is supplied
electric motors or gear trains turning at significantly lower from an external source and the bearing support forces are not a
speeds and for shorter times than are of interest here, so these first order function of speed, the bearing is termed hydrostatic.
made do with dry friction bearings operating for limited peri- When the bearing support forces are derived from the motion of
ods. The higher speeds and longer lives desired for micro-heat the rotor, then the design is hydrodynamic. Hybrid implemen-
engines require low friction bearings. The very small size of tations combining aspects of both are also possible. Since the
these devices precludes the incorporation of commercially avail- MEMS gas turbines include air compressors, both approaches are
13 Copyright ©2003 by ASME
Diameter (mm) x RPM yields taper ratios of about 30:1 to 50:1 on narrow (10-20 µm)
9 ×10
6
1000k
0 1 2 3 4 5 6 7 8 etched vertical channels 300-500 µm deep [15]. This capability
L/D = 1.0 defines the bearing length while the taper ratio delimits the bear-
Conventional ing gap, g. For hydrodynamic bearings we wish to maximize the
Load Capacity per Unit Area (Pa)
foil bearing
footprint and minimize gap/diameter to maximize load capac-
ity, so the bearing should be on the largest diameter available,
100k L/D = 0.25 the periphery of the rotor. The penalty for the high diameter is
relatively high area and surface speed, thus high bearing drag,
Hydrodynamic
microbearings L/D = 0.075 and low L/D and therefore reduced stability. In the radial 4000-
µm-diameter turbine shown in Figure 6, the journal bearing is
10k 300 µm long and about 15 µm wide, so it has an L/D of 0.075,
L/D = 0.075
g/D of 0.038, and peripheral Mach number of 1. This relatively
Hydrostatic microbearing short, wide-gapped, high-speed bearing is well outside the range
of analytical and experimental results reported in the gas bearing
1k literature. It is much closer to an air seal in aspect ratio.
The dynamical behavior of the rotor is of first order con-
0 100 200 300 400 500 cern because the high rotational speeds needed for high power
Tip Speed (m/s)
(Courtesy of L. Liu) density by the turbo and electrical machinery require the rotor to
operate at rotational frequencies several times the lowest radial
Figure 18: Gas bearing radial unit load capacity variation resonant frequency of the bearing/rotor system. The dynamics
with speed. of gas bearings on a stiff rotor can be simply represented by the
rotor mounted on a set of springs and dampers, as illustrated in
applicable. Both can readily support the loads of machines in this Figure 19. The fluid in the bearing acts as both the springs and
size range and can be used at very high temperatures. The two the principal source of damping. It also generates the destabiliz-
types of bearings have differing load and dynamic characteristics. ing cross-stiffness forces which cause instability at high speeds.
In hydrodynamic bearings, the load capacity increases with the As in many conventional engines, the rotor must traverse the
speed since the film pressure supporting the rotor is generated critical frequency and avoid instabilities at higher speeds. For
by the rotor motion. This can be true for a hydrostatic bearing as example, Figure 20 illustrates the whirl radius versus speed for
well if the film pressure is increased with increasing rotor speed, a 4-mm-diameter turbine with a 12-µm-wide bearing. Plotted
for example if the pressure is derived from an engine compres- on the figure are experimental data and a fit of an analytical
sor. However, when the supporting film pressure in a hydrostatic fluid mechanic spring-mass-damper model of the system to
bearing is kept constant, the load capacity decreases slightly with that data. The resonant peak amplitude is reached as the rotor
increasing speed. The calculated unit load capacity (support force crosses a “rotor critical” (resonant) frequency. If the peak excur-
per unit area of bearing) of plane journal microbearings is com- sion exceeds the bearing clearance, then the rotor hits the wall,
pared with the measured capacity of conventional air foil bearings i.e. “crashes”. A well-known characteristic of a spring-mounted
in Figure 18. The hydrostatic bearing is at a constant pressure. For rotor system (a so-called “Jeffcott rotor”) is that at speeds below
hydrodynamic bearings the load capacity is a function both of the critical frequency the rotor revolves around its geometric
rotational speed and of bearing length (L) to diameter (D) ratio. center, while well above the critical frequency the rotor revolves
Microbearings currently have low L/D’s due to manufacturing around its center of mass. Thus the dotted line in the figure, the
constraints, so their load capacity is less.
The relevant physical parameters determining the bearing
behavior are the length-to-diameter ratio (L/D); the journal gap- ∆P
to-length ratio (g/L); and nondimensional forms of the periph-
eral Mach number of the rotor (a measure of compressibility), Bearing
the Reynolds number, and the mass of the rotor. For a bearing gas flow
supported on a hydrodynamic film, the load bearing capability k ω
scales inversely with (g/D)5 which tends to dominate the design Rotor
m D
considerations [41]. g
The design space available for the micro-journal bearing is Rotor
greatly constrained by manufacturing capability, especially if the
rotor and journal structure are fabricated at the same time (which
avoids the need for assembly and so facilitates low cost, wafer- (Not to scale)
L
level manufacturing). The most important constraint is the etch-
ing of vertical side walls. Recent advances of etching technology Figure 19: Gas journal bearing model.
14 Copyright ©2003 by ASME
12 mity is a superior approach.
Data Hydrostatic bearings are stable from zero speed up to the
Model
Rotor Radial Excursion (µm)
10 stability boundary. However, centered hydrodynamic bearings
are unstable at low rotational speed but stable at high speeds.
8
Commonly, such bearings are stabilized by the application of a
unidirectional force which pushes the rotor toward the journal
6
wall, as measured by the eccentricity, the minimum approach
4 distance of the rotor to the wall as a fraction of the average gap
(0 = centered, 1 = wall strike). At conventional scale, the rotor
2 weight is often the source of this side force. At microscale, (1)
Rotor Imbalance the rotor weight is negligible, and (2) insensitivity to orienta-
0 tion is desirable, so a scheme has been adopted which uses
0 20,000 40,000 60,000 80,000
Rotational Speed (rpm) differential gas pressure to force the rotor eccentric. Extensive
(Courtesy of C.J. Teo) numerical modeling of these microbearing flows has shown that
such a rotor will be stable at eccentricities above 0.8-0.9 [45].
Figure 20: Transcritical response of the micro-journal gas For the geometry of the turbine in Figure 6, the rotor must thus
bearing in Figure 6. operate between 1-2 µm from the journal wall. This implies that
deviations from circularity of the journal and rotor must be small
asymptote of the curve fit, is a measurement of the rotor imbal- compared to 1 µm, an additional manufacturing requirement.
ance expressed in terms of radial displacement of the rotor center A rotor must be supported against axial as well as radial
of mass from the geometric center. The measured imbalance loads and so requires thrust bearings in addition to the radial
shown in the figure is ~2 µm, compared to the 12 µm bearing bearings discussed above. Both hydrostatic and hydrodynamic
clearance (i.e. at 12 µm imbalance the rotor would strike the wall approaches have been demonstrated. In either approach, the
on every revolution). bearing must support the axial loads and remain stable. The
We thus have two rotor dynamic design considerations, devices built to date have been designed for sub-critical thrust
traversing the critical frequency and ensuring that the frequency bearing operation so that bearing behavior traversing the critical
for the onset of instability is above the operating range. For a frequency is not an issue.
hydrostatic bearing the critical frequency simply scales with the Hydrostatic thrust bearings meter external air through
pressure in the bearing. The damping ratio (mainly viscous damp- supply orifices onto the bearing surface. The 400-µm-diameter
ing) decreases with increasing speed. Thus, the maximum ampli- thrust pad at the rotor center of the 4-mm-diameter turbine in
tude the rotor experiences while crossing the critical frequency Figure 6 rotates relative to a stationary thrust bearing surface
increases with bearing pressure, i.e. the peak in Figure 20 moves of similar diameter. The stationary bearing surface is perforated
up and to the right with increasing pressure [42]. This suggests with a circular array of 12-µm-diameter nozzle orifices fed from
the strategy of crossing the critical frequency at low pressure and a plenum which supplies the gas lubricating film between the
low speed and then increasing the pressure to stiffen the bearing bearing surfaces. A cross-section is shown in Figure 21. At a
as the rotor accelerates to increase the speed at onset of instability rotor-stator gap of 1.2 µm, a flow of 10 sccm at 2-5 atm is needed
[43]. to provide sufficient load capacity (0.5 N) and axial stiffness
The rotor imbalance is another factor which influences both
the peak amplitude crossing the critical frequency and the onset Spiral Bearing
Grooves Pressure
of instability. Large rotating machines are usually dynamically Rotor Rotor
balanced by measuring the imbalance and then adding or sub-
tracting mass to reduce it. In many micromachines it is possible Stator Stator
to avoid the need for dynamic balancing because the base mate-
rial used to date (single-crystal silicon) is extremely uniform
and, with sufficient care, the etching uniformity is sufficient
to produce adequately balanced rotors. Typically, the center of
mass is within 1-5 µm of the geometric center. For the turbine
in Figure 6, the blades must be etched to about ±1 µm span uni- Center
formity across the 4000-µm-diameter disk. For rotors made up Vent
of several wafers, the alignment between wafers must also be a) Hydrodynamic b) Hydrostatic
considered (also about 1 µm is needed) [44]. Using the balance Thrust Bearing Thrust Bearing
measurement capability evident in Figure 20 and laser etching, it
is also possible to dynamically balance a microrotor. It is unclear Figure 21: Geometry of (a) hydrodynamic and (b)
at this time whether dynamic balancing or manufacturing unifor- hydrostatic thrust bearings (not to scale).
15 Copyright ©2003 by ASME
Table 2: Design Considerations and Material Properties of
Interest for Gas Turbines
Orifice
Spiral
Grooves
Figure 22: Hybrid hydrodynamic (spiral grooves) and
hydrostatic (orifices) 0.7 mm dia thrust bearing. STRUCTURES AND MATERIALS
Structural considerations for the design of a MEMS gas tur-
bine are in many ways similar to those of conventional engines.
(2×105 N/m). Stiffness is maximized when the pressure drop The design space is defined by the requirements of the thermo-
through the supply orifices equals that of the radial outflow from dynamics (which require high stress and high temperatures), the
the orifice discharge to the bearing edge. properties of the materials, and the manufacturing capabilities.
Hydrodynamic thrust bearings use viscous drag, often The material properties, in turn, are very much dependent on
enhanced with shallow spiral grooves, to generate a pressure their processing. This section reviews materials selection, struc-
gradient in the bearing which increases toward the rotor center. tural design features, high temperature structures, analysis of
The pressurized gas film provides the bearing load capacity such microstructures, and packaging (installation) technology.
and stiffness. This self-pumping eliminates the need for an
external air supply and simplifies the manufacture since bear- Materials
ing air supply plumbing is not required, reducing the number Materials for gas turbine engines must exhibit high specific
of wafers needed. It also adds an additional design consider- strength (strength/density) at high temperatures. High tempera-
ation – rotor liftoff, i.e. the minimum rotational speed needed ture operation also requires creep and oxidation resistance. Other
to develop sufficient pressure to eliminate rubbing between the properties of interest include fracture toughness, modulus, and
stationary and rotating parts. Figure 22 is a 700-µm-diameter resistance to thermal shock (Table 2). MEMS processing tech-
hydrodynamic thrust bearing with 1.5 µm deep spiral grooves nologies are much more mature for silicon than for other materi-
that was tested on the microturbine of Figure 6. It lifts off at als so it is the first material a MEMS engineer considers (not so
about 80,000 rpm. Such a bearing at 106 rpm will dissipate for a gas turbine designer). In terms of strength at temperature,
about 0.2 watts, about the same as that of a hydrostatic bearing single-crystal Si is the equal of common nickel alloys and,
of equal load capacity [46]. because it has only 1/3 the density, its specific strength is much
higher, as illustrated in Figure 23. It is quite oxidation-resistant
103
Ultimate Tensile Yield Stress (MPa)
CVD SiC
Non-Dim. Thermal Stress (/K)
10-2
(Strength/Density) (m/sec)2
Inconel 600
CVD SiC
Specific Yield Stress
Hastelloy X
SiC
Inconel 600 105 Hastelloy
X
102 Hastelloy X
Si
10-3
Inconel 600
Si 104 Si
101
200 600 1000 1400 200 600 1000 1400 200 600 1000 1400
Temperature (K) Temperature (K) Temperature (K)
Figure 23: Material properties relevant to high speed, high temperature rotating machinery.
16 Copyright ©2003 by ASME
and has thermal conductivity approaching that of copper, so it 800
is resistant to thermal shock. On these grounds, it is not a bad σfracture, SiC = 1.0 GPa
material for gas turbine engines. However, at temperatures % SiC dε/dt = 1E-4 sec-1
below about 900 K, Si is brittle, so usable strength is very much
a function of the details of the processing. Structural life must be 600 30
Si
Allowable Stress, σ (MPa)
assessed with statistical methods. 20 σ SiC σ
Chen et al. [47] have reported room temperature strengths Si
up to 4 GPa for micromachined Si specimens. Moon et al. [48] 10
have measured the strength and creep rate of Si at temperatures 400
up to 1000 K. From these measurements and a detailed model
of the creep behavior of the material, it appears that long-lived
structures can be designed for stress levels up to about 500 MPa
at 850 K. Oxidation is another concern for high temperature 200 0
structures. Conductively-cooled Si combustor tests were run at
exit temperatures up to 1800 K [33]. The thickness of the oxide
layers grown were in agreement with standard models of Si oxi-
dation. These imply that uncoated Si airfoils can have a life of a
0
few hundred hours. Longer lives may require coatings. Si nozzle 600 700 800 900
guide vanes run for 5 hrs at 1600 K in a microcombustor exhaust Temperature (°C)
show little degradation, Figure 24. (Courtesy of H-S Moon)
Silicon carbide has about 600 K more temperature capabil-
ity than Si, but the SiC microfabrication technology is much Figure 25: Usable strength of Si/SiC/Si hybrid structure in
tension.
less mature. SiC is available in single-crystal wafers and can
be precision-etched but SiC wafers cost 100x more than Si at
the moment and etch rates are about 10x slower. An alternative reaction sintering of powdered SiC to form parts such as turbine
to direct SiC etching is to etch a female mold in Si and then fill rotors [50].
the mold (for example by chemical vapor deposition, CVD), and Additional structural materials of interest for MEMS gas
dissolve away the Si, leaving an SiC precision structure. The turbines include glasses for thermal and electrical isolation,
challenges here are realizing SiC with the needed mechanical and very high temperature materials such as sapphire. There is
properties and dealing with the intrinsic stresses induced by the considerable microfabrication experience with glass but very
combination of the high temperatures of the CVD process and little with the refractory ceramics because these have not been
the difference in coefficient of thermal expansion between the considered as MEMS materials in the past.
two materials. A variation on this approach is to use CVD to
fill cavities in Si wafers with SiC, bond another Si wafer over Structural Design Considerations
the filled cavity, and then process the pair as though it were a Structural design of a MEMS gas turbine has many of
standard Si wafer. This yields SiC-reinforced silicon structures the same considerations as the design of large machines: basic
which have more temperature capability than Si but are easier to engine layout is set by rotor dynamic considerations, centrifugal
manufacture than SiC [49]. The increased temperature capabil- stress is the primary rotor load, stress concentrations must be
ity of a turbine like that in Figure 6 increases with the thickness avoided, and hot section life is creep- and oxidation-limited.
of the SiC insert, Figure 25. Another approach being pursued is Some large engine concerns do not exist at micron scale. For
example, the microrotors are very stiff so that backbone bending
is not a concern; thermal stress from temperature gradients is not
important at these sizes; maintenance is not a design issue; and
fasteners do not exist here so the engineering details involved
with bolting, static sealing, etc. do not exist [51].
Although many of the design considerations are independent
of size, the engineering values are not, of course. Airfoils need fil-
lets at the roots to avoid stress concentrations with radii of 10-30
µm. Surface finish is important with roughness measured in nano-
meters. Forced response excitation of blade rows must be avoided
with blade-bending frequencies on the order of megahertz rather
than kilohertz, the rotor once-per-rev frequency is 20 KHz rather
Figure 24: 200 μm high, Si turbine blades new and after 5 hrs than 200 Hz. For the turbine of Figure 6, the lowest blade mode is
at 1600 K gas temperature in a microcombustor exhaust. 2.5 MHz while the blade passing frequency is 0.9 MHz.
17 Copyright ©2003 by ASME
Below 850-900 K, silicon is a brittle material so that proba- literature on MEMS sensors and valves, literally thousands of
bilistic analysis is a preferred method for failure analysis. Such papers, and some units are commercially available. As for large
techniques applied to the turbine rotor geometry of Figure 6 at engines, however, the combination of harsh environment, high
peripheral speeds of 500 m/s predict failure probabilities of 10-10 frequency response, high accuracy, and high reliability means
to 10-8, depending upon the flaw population assumed [52]. In a that sensors and actuators for MEMS-scale gas turbines must be
rotor constructed from single-crystal Si, all of the flaws are likely specifically engineered for that environment.
to be surface flaws. In a rotor of CVD SiC, volumetric flaws may Engine control laws are generally based on reduced order
also exist. In either case, the flaw population and thus the usable models of the engine dynamics. The dynamics of millimeter-
strength of the material is a strong function of the manufacture, scale engines are, of course, much faster than larger engines and
as it is at any scale. A variation of a factor of four in strength can also include phenomena not seen in large engines. The addi-
has been reported for deep-etched Si depending upon post-etch tional dynamics arise if there is significant heat transfer from the
surface treatment [53]. hot section into the compressor [55]. In this case the heat transfer
Large engines use standard tubing fittings and electrical degrades the compressor performance so that the pressure ratio
connectors to pass fluids and electrical signals to the outside and mass flow are a function of hot section temperature as well
world. These do not exist at microscale. Most computer chips as shaft speed. Since the heat transfer has a time constant not
and MEMS devices do not require fluid connections and those much faster than the rotor acceleration, it alters the dynamics of
that do operate not much above room temperature. For these the gas turbine from that of a first order system (as large engines
applications, there are a variety of adhesives and polymer are) to a second order system, requiring additional sophistica-
systems. A micro-gas turbine can have a surface temperature tion in the control law design. The best way of avoiding this
above 700 K and require fluid connection at pressures of 10 complexity is to thermally isolate the hot and cold sections of
atm or more, so that high strength, high temperature packaging the engine, which is, of course, desirable for improved thermo-
approaches are needed. One approach which has proven success- dynamic performance.
ful is an adaptation of the hermetic package technology used for
military electronics. This joins Kovar (a nickel alloy) tubing to Sensors
silicon using glass as the bonding agent. The joining is done in Large-scale gas turbines use compressor pressure ratio and/
a furnace above 1100 K to melt the glass. Such joints can with- or rpm as the primary input to the fuel control system. Sensor
stand pressures above 200 atm [54]. selection for a MEMS engine is a trade among observability of
the state (dynamical information represented by the sensed quan-
ENGINE CONTROLS AND ACCESSORIES tity), response time, difficulty of fabrication, and environmental
All gas turbine engines require control systems to insure compatibility. Liu [55] used a dynamic model of a MEMS engine
safe operation. Typically, the control system adjusts the fuel flow to evaluate the suitability of various sensor options including
to deliver the requested power, and monitors engine operation to rpm and compressor discharge pressure or temperature. Of these,
avoid unsafe conditions such as over-speed, over-temperature, rpm is the most sensitive and temperature the least. Sensing is
or surge. Such a control system consists of sensors (speed, pres- complicated by the high rotational frequencies (1,000,000 rpm)
sure, temperature, etc.), a feedback controller with a suitable set and high temperatures (600 K at the compressor discharge) in a
of control laws (now implemented in a digital computer), actua- very small engine. In principle, sensors can be fabricated inte-
tors such as a fuel control valve (often called a fuel management grally with an engine or located remotely, with each approach
unit or FMU), and compressor stability devices as needed (bleed presenting challenges. A sensor remote from the gas path suffers
valves, fences, variable stators). Engine accessories include an reduced frequency response, which is already a challenge at the
ignition system, fuel pump, lubrication system, and starter. All MEMS scale. In addition, it is difficult to be very remote from
of this functionality is needed for a millimeter-scale MEMS gas the gas path in an 0.5-mm-long engine. Integral sensors must
turbine and all must fit within a micro chip if the accessories withstand the high temperature of the gas path. Even for sensors
are not to dwarf the engine. Following in the tradition of large fabricated in the cold sections, they must be capable of with-
engine development, the controls and accessories have received standing a wafer-processing environment that exceeds 1000-
less attention to date than the major engine subsystems such 1400 K. This effectively precludes the use of low temperature
as the compressor, turbine, and combustor but they are no less materials such as polymers and most metals in the device design.
important to the ultimate success of the concept. Integral sensors have several advantages, however. They can be
very small and thus have high frequency response, and many can
Engine Controls be fabricated in parallel for low cost redundancy.
The simplest engine control would consist of a single sensor One integral solution was developed by Tang [56] who
feeding a digital controller which commands the fuel flow rate adapted a hot-film-type sensor to this application (Figure 26).
valves. The functional requirements for the valving and the sen- Designed for placement on the wall above a rotor blade tip, this
sors stem from the engine dynamics as represented in the control 50 µm square sensor is a heated, serpentine, polysilicon resistor
laws and from the engine environment. There is a very rich positioned over a trench for thermal isolation. The sensor and its
18 Copyright ©2003 by ASME
flows 35 sccm of N2 at a pressure drop of 0.5 atm. Frequency
response is several hundred hertz. Cyclic testing of the valve
has demonstrated a 50,000+ cycle life for the units tested. This
design is on-off. The actuation-pressure scaling laws favor small
valves so the intent is to use a parallel array of 20 on-off valves,
each with a capacity of 5% of the maximum fuel flow, to meter
the fuel. All 20 valves would consume less than 1 mW total
power and operate at temperatures approaching 1000 K so they
can be embedded in an engine chip. Many other arrangements
are possible, such as logarithmic spacing of the valve orifice
sizes to give finer fuel flow control.
Starter-Generator
Microelectrical machinery is required for power generation
and electric starting, if desired. There is an extensive literature
on microelectric motors, which is not reviewed here, but little
Figure 26: A 50 μm sq hot film RPM and temperature sensor. work on generators. The requirements for the devices of interest
here differ from previous work in that the power densities needed
are at least two orders of magnitude greater than that of con-
leadouts are all polysilicon which is selectively doped to adjust ventional-size and previous micromachines. Also, the thermal
its resistivity (high for the sensor, low for the leads). Polysilicon environment is much harsher. Integrating the electric machine
has the advantage that it is compatible with most semiconduc- within the engine offers the advantage of mechanical simplicity
tor fabrication techniques and can withstand high temperatures. in that no additional bearings or structures are required over that
Simulations confirmed by shock tube testing showed this needed for the fluid machinery. There is also a supply of cooling
approach to have sufficient sensitivity and frequency response air available.
to respond to the flow perturbations above a compressor blade Both electric and magnetic machine designs can be con-
tip as predicted by a 3-D CFD simulation. With a total thickness sidered and, to first order, both approaches can yield about
of less than 1 µm, such sensors could be fabricated on the casing equivalent power densities. Since the magnetic machines are
above the compressor blade tips. This type of resistor has also material property-limited at high temperature and because of
been shown to be usable as an igniter. the challenges of microfabricating magnetic materials (which
are not compatible with standard semiconductor manufacturing
Fuel Control Valves techniques), electric designs were first explored. Power density
Very small engines are the topic of this discussion so the fuel scales with electric field strength squared, frequency, and rota-
control valves should be equivalently small. If integrated within
the engine, the valve design must then be fully compatible with
the fabrication and operating conditions of the gas turbine. This
choice strongly constrains the valve design space. For example,
the high processing and operating temperatures prohibit the use
of polymers, so a hard valve seat must be used. The principal
design requirements are flow rate, pressure, frequency response,
very low power consumption and leakage, and high temperature
capability.
Yang et al. [57] developed MEMS fuel-metering valves
for gaseous hydrocarbon fuels such as propane. The design is a
simple silicon spring-mounted plunger opened by electrostatic
forces and closed by a combination of the spring and fluid
pressure forces. The electrostatic approach has the advantages
that very little power is needed to open a valve (40 nW) and
the electrical materials (polysilicon) are compatible with high
temperature semiconductor fabrication technology. Such a valve
is shown in Figure 27. The 2-mm-square valve has a 1000-µm-
diameter plunger which rises 3 µm off the seat when actuated.
An 8-inch wafer of valves would contain about 5000 individual
units. The valve opens against 10 atm pressure and, when open, Figure 27: A 1 mm dia fuel control valve on Si beam springs.
19 Copyright ©2003 by ASME
Stator Phase Interconnect
Substrate "Bus-Bar"
(mechanical (1 for each
support) ons Stator
micr
Electric Field Stator of 6 phases)
in Air Gap Stator 2000 Electrode
Insulator
2 m Stator
Electrodes
Rotor Film
Rotor Rotor Rotor
Motion Insulator
Rotor 4 m{
Substrate
(mechanical
support) Figure 29: A 131-pole, 6-phase, 4 mm dia electric induction
stator.
Figure 28: Fields and charges in a microscale electric
induction motor-generator.
To maximize power output, induction machines such as
these require the spacing between the rotor and stator to be
tional speed. The micromachinery of interest here operates at on the order of the stator pitch. The electrical torque produced
peripheral velocities 1-2 orders of magnitude higher than previ- scales with the square of the rotor-stator spacing, a few microns
ously reported micromotors, and so yields concomitantly more in this case. However, in these high-speed machines, the rotor
power. Electric machines may be configured in many ways. Here periphery is at sonic velocity so the viscous drag in a gap of only
an induction design was chosen since it requires neither electri- 2-3 microns is extremely large. Indeed, this drag is the major loss
cal contact with the rotor nor knowledge of the rotor position. mechanism for such an electrical machine. Thus, there is a basic
The operation of an electric induction machine can be under- design tradeoff for the electric motor-generator between power
stood with reference to Figure 28 [58]. The machine consists of density and efficiency. While it may be possible to alter the local
two components, a rotor and a stator. The rotor is comprised of geometry to reduce the drag somewhat [24], the drag still makes
a 5-20 µm thick good insulator covered with a few microns of a up about half the total loss and limits the efficiency (shaft to net
poor conductor (200 MΩ sheet resistivity). The stator consists of electrical) of these designs to 40-50%.
a set of conductive radial electrodes supported by an insulator. A magnetic induction machine has many fewer poles so that
A traveling electric potential is imposed on the stator electrodes the optimum rotor-stator spacing is much larger (30-50 µm) and
with the aid of external electronics. The resulting rotating elec- the drag concomitantly lower. Koser and Lang [61] designed
tric field then induces an image charge on the rotor. Depending such a machine based on the microplating technology devel-
on the relative phase between the motion of rotor charges (set by oped by Park et al. [62]. A four-pole stator for a 4-mm-diameter,
the rotor mechanical speed) and that of the stator field (set by the induction motor-generator, designed to be functionally equiva-
external electronics), the machine will operate as a motor, gen- lent to the electric machine in Figure 29, is shown in Figure 30.
erator, or brake. Torque increases with the square of the electric More recent versions of this stator includes a laminated magnetic
field strength and frequency. The maximum electric field strength
that an air gap can maintain without breakdown is a function of
the gap dimension. In air, the breakdown field is a maximum at a
gap of a few microns so that micromachines can potentially real-
ize higher power density than large machines of the same design.
Frequency is constrained by external electronics design and by
fabrication constraints on the stator electrode geometry. Current
technology is limited to about 300 volts and 1-2 MHz. This is
consistent with a 6-mm-diameter machine producing about 10
watts at a 3 µm air gap. A 4-mm-diameter, six-phase, 131-pole
(786 electrodes, each 4 µm wide) stator for such a machine is
shown in Figure 29 [59]. Note that such an electric motor-gen-
erator occupies less than 20 µm thickness at the surface of the
rotor and stator (mainly the insulator thickness). Thus the power
density of this machine (excluding the external electronics) is
many times that of a conventional magnetic motor-generator, on
the order of 100 MW/m3. Fréchette et al. [60] have reported a
(Courtesy of M. Allen)
similar design run as an electric motor. The torque produced by
these devices has agreed with theoretical predictions but high Figure 30: A 4-pole stator for a 4 mm dia magnetic motor-
power operation has yet to be reported. generator.
20 Copyright ©2003 by ASME