A robust micro conveyer realized by arrayed polyimide jointactuators
ABSTRACT A new micro motion system (micro-conveyer) based on arrays of movable robust silicon legs has been developed and investigated. The motion is achieved by thermal actuation of polyimide joint actuators using electrical heating. Successful experiments on moving flat objects in the millimeter range with high load capacity have been performed. The conveyer consists of a 15×5 mm2 chip with 12 silicon legs each 500 μm long. The maximum load conveyed on the structure was 2000 mg. Conveyance velocities up to 12 mm/s have been measured. Accelerated lifetime measurements demonstrate the long-term stability of the actuators. The function of the polyimide joint actuators is unaffected after more than 2×108 load cycles
- SourceAvailable from: Sjoerd Haasl
Conference Proceeding: Robust, large-deflection, in-plane thermal polymer V-shaped actuators[show abstract] [hide abstract]
ABSTRACT: We present the design, fabrication and characteristics of the first in-plane V-shaped polymer actuator. Several different designs are evaluated and compared with each other. The intended application for this actuator is optical component alignment.Micro Electro Mechanical Systems, 2004. 17th IEEE International Conference on. (MEMS); 02/2004
- [show abstract] [hide abstract]
ABSTRACT: This paper describes electrothermal microactuators that generate rectilinear displacements and forces by leveraging deformations caused by localized thermal stresses. In one manifestation, an electric current is passed through a V-shaped beam anchored at both ends, and thermal expansion caused by joule heating pushes the apex outward. Analytical and finite element models of device performance are presented along with measured results of devices fabricated using electroplated Ni and p<sup>++</sup> Si as structural materials. A maskless process extension for incorporating thermal and electrical isolation is described. Nickel devices with 410-μm-long, 6-μm-wide, and 3-μm-thick beams demonstrate 10 μm static displacements at 79 mW input power; silicon devices with 800-μm-long, 13.9-μm-wide, and 3.7-μm-thick beams demonstrate 5 μm displacement at 180 mW input power. Cascaded silicon devices using three beams of similar dimensions offer comparable displacement with 50-60% savings in power consumption. The peak output forces generated are estimated to be in the range from 1 to 10 mN for the single beam devices and from 0.1 to 1 mN for the cascaded devices. Measured bandwidths are ≈700 Hz for both. The typical drive voltages used are ⩽12 V, permitting the use of standard electronic interfaces that are generally inadequate for electrostatic actuatorsJournal of Microelectromechanical Systems 07/2001; · 2.13 Impact Factor
IEEE MEMS’99, Orlando, USA, Jan 17-21, 1999
A ROBUST MICRO CONVEYER REALIZED BY
ARRAYED POLYIMIDE JOINT ACTUATORS
Thorbjörn Ebefors *, Johan Ulfstedt Mattsson, Edvard Kälvesten and Göran Stemme
Department of Signals, Sensors and Systems, Royal Institute of Technology,
SE-100 44 Stockholm, Sweden*E-mail: Thorbjorn.Ebefors@s3.kth.se
A new micro motion system (micro-conveyer) based on arrays of movable robust silicon legs has been
developed and investigated. The motion is achieved by thermal actuation of polyimide joint actuators using
electrical heating. Successful experiments on moving flat objects in the millimeter range with high load
capacity have been performed. The conveyer consists of a 15x5 mm2 chip with 12 silicon legs each
500 µm long. The maximum load conveyed on the structure was 2000 mg. Conveyance velocities up to
12 mm/s have been measured. Accelerated life-time measurements demonstrate the long-term stability of
the actuators. The function of the polyimide joint actuators is unaffected after more than 2x108 load cycles.
When new products become smaller and smaller, e.g. by using
MEMS techniques, requirements on assembling the small sub-
systems in large complex systems require new assembling tools.
The same is true for a more futuristic approach concerning
micro-factories which requires automated handling and assembly
of small parts with accuracy in the sub-micron range.
Miniaturized micro-motion systems such as micro-conveyers and
micro-robots fabricated with MEMS-techniques can be a part of
the solution to these handling and assembling problems. Further
applications for micro motion systems might be found in the IC
manufacturing industries which require accurate moving,
positioning and manipulation of small parts.
During the last years a variety of MEMS-concepts for
realization of micro motion systems in the form of micro-
conveyers have been presented [1-11]. Some of them are
summarized in table 1. The micro-conveyers can be classified in
two groups: contact free (CF) or contact (C) systems, depending
on whether the conveyer is in contact with the moving object or
Contact free (CF) system have been realized using pneumatic,
electrostatic or electro-magnetic forces to create a cushion for the
mover to levitate on. Magnetic levitation can be achieved by
using either permanent magnets, electromagnets or diamagnetic
bodies (i.e. a superconductor) . The main advantage of the
contact free systems is low friction. The drawback is that these
system are very sensitive to the cushion thickness (i.e. load
dependent) and have low load capacity. The cushion thickness
can also be quite difficult to control.
Table 1. Overview of some micro-motion systems.
(C=Contact, CF=Contact Free)
Length per stroke/
Pister et al.,
CF: pneumatic air bearing (for low-
friction levitation) + electrostatic force for
flat Si pieces
< 1,8 mg
max at 1-2 Hz
Fujita et al,
CF: magnetic levitation (Meissner effect)
+ magnetic Lorentz force for driving
7,1 mm/s a
Nd-Fe-B magnet slider
not specifiednot specified
Fujita et al,
C: array of thermobimorph polyimide legs
(cantilevers); electrical heating (asyn)
flat Si piece
∆x= 80 µm (f<fc; 33mW)
f =1 Hz (pressure)
8 x 2 x 16 legs á 500 µm
total area: 5 x 5 mm2
Fujita et al,
CF: array of pneumatic valves;
flat Si piece
9 x 7 valves á 100x200 µm2
total area: 2 x 3 mm2
Ho et al,
C: array of magnetic flap actuators.
External magnet for actuation (syn)
2.6 mm/s b
flat Si pieces
< 222 mg
∆x= 500 µm
∆x= 5 µm
fc high kHz-range
∆x= 20 µm
fc not specified
4 x 7 x 8 flaps á 1400 µm
total area: 10 x 10 mm2
et al, 1996 [6, 7]
C: array of torsional 5µm high
Si-tips; electrostatic actuation (asyn)
flat glass piece
≈ 1 mg
15000 tips 180 x 240mm2
total area: 1000 mm2
et al, 1997 
C: array of thermobimorph polyimide legs;
therm. + el.static actuation (asyn)
8 x 8 x 4 legs á 430 µm
total area: 10 x 10 mm2
Yonezawa et al,
CF: array of planar electromagnets 30 mm/s c
(110 mg magnet)
35 mm/s d
(for flat objects)
flat magnet +
ext. load <1200 mg c
≈40 x 40 coils á 1 x 1 mm2
total area: 40 x 40 mm2
de Rooij et al,
CF: Pneumatic (air jets) sliders of Si
< 60 mg
2 x 10 slits á 50 µm
total area: 20 x 30 mm2
C: array of erected Si-legs; thermal
actuation of polyimide joints (asyn)
12 mm/s e
flat Si pieces +
ext. load 2100 mg
∆x=170 µm e (f<fc; 175mW)
fc= 3 Hz f
2 x 6 legs á 500 µm
total area: 15 x 5 mm2
aThe superconductor requires low temperature (77 K).
bEstimated cycletime ≈ 25 ms (faster excitation results in uncontrolled
jumping motion) and 0.5 mm movements on 8 cycles .
cDepends of the magnet and surface treatment.
d For flat sliders. The velocity depends of the surface of the moving slider (critical
tolerances of the slider dimensions).
ePossible to improve with longer legs and more V-grooves in the joint.
fPossible to improve. Thinner legs with smaller polyimide mass to heat would
Different kinds of systems where the actuators are in contact (C)
with the moving object have been realized based on arrays of
movable legs erected from the silicon wafer surface. The legs
have been actuated by using different principles such as thermal,
electrostatic and magnetic actuation. Both synchronous driving
 and the more complex but also more effective asynchronous
driving mode [3, 11] have been used to realize this kind of
system. Asynchronous driving techniques require at least two
spatially separated groups of actuators which are turned on and
off at different times, holding and driving the object alternately.
High speed and smoother motion can be achieved with such
The challenge with the micro-motion systems is to realize
enough load capacity to meet the requirement of moving and
positioning for example silicon wafers, electronic chips,
multiple chip modules (MCM) and read/write heads for hard discs
with typical weights of the order of hundreds of mg/cm2. The
main drawback of some of earlier presented micro-conveyers
[3, 6, 7, 10] is that they used surface micromachined fabrication
techniques which results in thin fragile legs and therefore
relatively low load capacity.
The presented micro-motion system based on electrostatic and
piezoelectric actuation has the advantage of low power
consumption but has a low load capacity. The solution is to use
magnetic or thermal actuation principles which are well known
for their potential to exert large forces and displacements. On the
other hand these actuation principles rely on significant currents
and/or powers which may require cooling. During the device
design one therefore has to deal with trade-off between range of
motion, strength, speed (actuation
consumption, control accuracy, system reliability, robustness,
load capacity etc. The magnetic actuation principle  has a
disadvantage due to its need of a magnet mover (slider) which
limits its usefulness. With a contact system based on thermal
actuators it is possible to move objects of various kinds (non-
magnetic, non-conducting, non-patterned, non-structured etc.).
The thermal principle also scales in a favorable manner (faster
and stronger) when going to micro-scale.
To achieve high performance micro-motion systems it is
important to overcome the drawbacks of the planar nature of
standard IC and micromachining fabrication techniques. Therefore
a variety of techniques based on out-of-plane rotation or folding
of silicon microstructures (like Japanese Origami) have been
proposed. An overview of such techniques is found in .
Pister et al. used a micro-hinge to produce leg components for
articulated micro-robots . Miura et al.  presented low
friction hinges based on polyimide to achieve 3-D folding of
small microstructures for mobile mechanism applications. The
flexibility of polyimide makes it a good micro-joint material.
Allen et al.  demonstrated an other type of actuator based on
polyimide joints using electrostatic actuation. Such a solution is
quite difficult to implement in array configurations.
We have recently developed a new robust polyimide micro-joint
useful both for static application such as 3-D sensors [16, 17]
and for dynamic applications [18, 19]. Our solution differs from
the other presented polyimide joints [14, 15] by being self-
assembled and having a strong actuator included inside the joint.
This makes it possible to fabricate arrays of individually
controlled silicon legs suitable for realization of micro-conveyers
with high load capacity.
This paper describes the principle, design, fabrication and
experimental performance of the new micro-conveyer. The final
goal is to realize a 2-D system for chip and wafer handling.
Polyimide joint actuator
The principle of the polyimide joint actuator is shown in
figure 1 and further described in published articles [16, 18, 19].
The legs are rotated out-of-plane (the self-assembled static mode)
using thermal shrinkage of the polyimide in the V-grooves
during the curing (imidization) process of the polyimide.
Polysilicon heaters are integrated in the polyimide joint. By
driving a heating power through the heaters local heating is
achieved in the joint resulting in thermal expansion of the
polyimide (dynamic mode). The V-shape of the joint results in
larger expansion at the top of the V-groove than at the bottom
which gives the dynamic motion, ∆x.
Principle of the micro motion system
Living organism offer good models for designing micro motion
systems [3, 20]. Mimicry of the way six legged insects walk
has been proposed for the design of multilegged robots possible
to implement using microfabrication techniques . The first
proposed  and realized  MEMS contact transportation
system was based on the ciliary motion principle adopted from
nature. Due to its simplicity for implementation with MEMS-
technology the same principle is used in this work. The simple
on-off actuator control can easily be realized with saturated FET-
transitors eliminating the need of onboard electronic on the
conveyer chip. The basic principle of the asynchronously driven
one dimensional micro conveyer is shown in figure .
The fabrication process is shown in figure 3. Here we used only
one poly-Si layer. Otherwise the process is the same as
previously described . The polyimide fabrication process is
compatible with standard CMOS processing  allowing the
integration of control circuitry for individual control of the
actuators in the array. Actuators with sensor functions and
Figure 1. Actuation principle for the leg movements based
on a four V-groove joint. By heating the joint a horizontal
displacement, ∆x, is obtained due to greater thermal expansion
of the polyimide at the top of the V-groove than at the bottom.
The silicon leg size is 500 x 600 x 30 µm3. The SEM-photo
shows a close-up of a fabricated five V-groove polyimide joint.
control circuitry on the same chip are of great importance for the
implementation of powerful manipulation strategies (e.g.
rotation and transportation movements in 2-D systems). figure 4
shows a photo of some of the fabricated (non-diced) micro-
conveyers and close-ups of the silicon legs.
Two kinds of basic experiments with single actuators and with
the whole conveyer system have been conducted. We started with
the characterization of single polyimide joint actuators which
also includes lifetime measurements. Then extensive experiments
concerning speed (actuation frequency), power consumption,
control accuracy, strength, robustness and load capacity of the
whole conveyer system were carried out.
Polyimide joint actuator characterization
Stroke-length and temperature measurements
The laser measurement set-up shown in figure 5 was used to
measure the horizontal displacement, ∆x, as a function of
supplied power and frequency for a four V-groove joint with a
static bending angle around 90°. A square wave heating current
results in a cut-off frequency, fc (- 3 dB) of 3 Hz, according to
figure 6. The maximum and minimum equivalent oscillation
temperatures in the polyimide during one cycle were calculated
for each frequency. This was obtained by comparing the bending
angle of the leg when the chip is heated at different temperatures
on a hot-plate with the bending angle obtained by electrical
heating. In this way a rough estimation of the polyimide
temperature was obtained. This temperature corresponds quite
well with the temperature obtained from measurements of the
resistance change of the polysilicon heaters. We have not been
able to do any direct temperature measurements to study the
temperature distribution inside the different V-grooves of the
joint. The inner polyimide filled V-grooves act as thermal
isolation giving the highest
displacement in the outermost V-groove.
temperature and largest
Figure 2. Operation principle of the asynchronous 1-D
micro conveyer [3, 11]. A displacement equal to 2·∆x is
obtained during one period due to the fixed phase difference of
90 degree between the two sets of legs (x+ and x- ). A 180
degrees phase-shift between x+ and x- results in movements of
the object in the opposite direction.
Silicon substrate with buried silicon dioxide
& thick resist
Figure 3. Schematic of the fabrication process based on SOI-wafers. The key steps are: (a) forming the integrated heater
using LPCVD-deposited polysilicon encapsulated in low-stressed silicon nitride and anisotropic KOH etching of 30 µm
deep V-grooves (b) local silicon dioxide (LOCOS) growth, via holes to the heaters, patterning of sputter deposited 1.5 µm
thick aluminum conductors (c) spinning and patterning of the polyimide in the V-grooves, a backside 500 µm KOH
silicon etch (d) dicing of the chip, a BHF oxide etch and solvent cleaning to release the chip from the protecting wafer and
also to release the 30 µm thick silicon legs. Finally a polyimide curing in an oven is done to erect the legs.
Figure 4. (a) Photograph showing different (non-diced)
structures used to demonstrate the function of the polyimide
joint based micro-conveyer. One conveyer consist of 2 rows
with legs, totally 12 silicon legs. The two set of legs (6 each
of x+ and x-) are indicated in the photo.
(b) SEM-photos showing Si-legs with a length of 500 µm.
A power of 90 mW/leg gives a temperature difference of
approximately 200°C during one actuation cycle and a stroke
length, ∆x of 100 µm (for low frequencies). For a 500 µm long
leg the largest displacement ∆x=170 µm was achieved at a
supplied power of 175 mW/leg. The maximum polyimide
temperature in the outermost V-groove is then above 400°C.
We have earlier described the dynamic behavior of the polyimide
joint using a simplified lumped heat capacity model . The
cut-off frequency is determined by the thermal mass, Cth, and the
thermal resistance, Rth of the polyimide joint. Polyimide joint
actuators with three V-grooves result in higher cut-off frequencies
due to the smaller polyimide volume to heat but also in smaller
stroke lengths, ∆x. To obtain fast cooling of the polyimide joint
the metal interconnections to the polysilicon heaters also are
used as heat conductors. With wide metal wires in the bottom of
the V-grooves we achieved good thermal conductance to the
silicon substrate (heat sink) and therefore fast cooling.
As figure 6 illustrates the temperature variations in the
polyimide are dependent on the frequency. At frequencies above fc
the thermal mass of the polyimide counteracts fast heating and
cooling which reduces both the maximum temperature and the
temperature variations. Therefore it is possible to compensate
the small displacements at higher frequencies by increasing the
heating power. Displacements, ∆x, larger than 40 µm for a four
V-groove joint have been achieved at 250 Hz using a heating
power of about 230 mW/leg. Such high heating power would
burn off the polyimide at lower frequencies, as will be described
in the next section.
Thermal failure mechanism
To test the maximum performance of the polyimide joint its
behavior at high polyimide temperatures (e.g. high power
dissipation) was studied.
At DC-powers typically above 180 mW/leg for four V-groove
joints and 200 mW/leg for three V-groove joints the outermost
V-groove with polyimide stops expanding and contracting when
it is heated by AC-power. When studying the (outermost) not-
actuated polyimide filled V-groove under a microscope the
polyimide looks burned and its color changes. The most likely
explanation of this ”burn off” is that the solvents inside the
polyimide outgas from the polyimide at high heating
temperatures (above 400°C). The polyimide starts to ”boil” and
crumple up. The polyimide adhesion to the silicon walls was
not affected. It is always the outermost V-groove that burns off
because that polyimide groove reaches the highest temperature.
Mechanical failure mechanism
Force measurements on the polyimide joint structures have been
presented earlier . For large forces (typical around
50-100 mN) and displacements of several hundred micrometers
the polyimide joint deforms plastically.
To further demonstrate the robustness of the polyimide joint
simple chock-test were performed on a single diced conveyer
chip. The chip with a weight of 75 mg was repeatedly dropped
from half a meter landing on its 12 silicon legs without
breaking either the polyimide joint or the silicon legs. Due to
the flexibility of the polyimide joints (Young’s modulus
E≈2-3 GPa) the erected silicon legs can be pressed down almost
to the wafer surface without breaking or deforming the
polyimide joint. Afterwards the silicon legs return to their
Lifetime measurements of the polyimide joint actuators were
carried out for five weeks using the measurement set-up shown
in figure 5. In order to have an accelerated test and conditions as
unfavorable as possible for the polyimide joint we used a
DC-offset in the power supply to obtain a high temperature of
approximately 200°C (well above room temperature) in the
polyimide joint. This DC-current minimized the possible
influence of bending angle caused by variation of the room
temperature during the measurements. By using an actuation
frequency of 2.5 Hz (just below the cut-off frequency) we
obtained the highest possible temperature difference in the
polyimide between on and off for a specific power as well as a
large number of load cycles.
The static bending angle increased by approximately 1.2° during
the initial stabilization phase during the first couple of hours.
This is probably caused by additional curing of the polyimide at
the high temperature. After the initial stabilization phase the
actuator gives stable bending angles for millions of cycles, as
can be seen in figure 7. The polyimide joint actuators worked
unaltered after more than 4.5 million load cycles (three weeks at
2.5 Hz) when the test was stopped.
l (1-3 m)
Detection of the
reflected laser beam
Figure 5. Measurement set-up for the bending angle, α,
and stroke length, ∆x.
0.11 10 100
displacement, ∆x [dB]
Equivalent polyimide joint
P= 90 mW/leg
max osc temp
min osc temp
Figure 6. Frequency dependence of the polyimide joint
temperature and the horizontal displacement, ∆x, for a
500 x 600 x 30 µm leg with a four V-groove joint.
The cut-off frequency, fc (- 3 dB) is 3 Hz. The predicted
minimum and maximum temperatures during the oscillation
cycle are also plotted.
With the intention of breaking the polyimide joint by fatigue the
actuator was run for an additional week at 250 Hz (>1.5x108
million load cycles). At frequencies well above the cut-off
frequency the maximum temperature in the polyimide is only
half of the maximum temperature obtained at low frequency.
Therefore we could expect breakage by fatigue and not by the
thermal failure mechanism described above where the polyimide
burns off. All four actuators used in the fatigue worked unaltered
after more than 2x108 million load cycles (250 hours at 250 Hz),
when the test was stopped, as can be seen in figure 7.
We have not seen anything that indicates that normal moisture
conditions affect actuator performance.
Velocity and load measurements
To evaluate the function of arrayed polyimide micro-joint
actuators basic experiments were performed with the different
prototype conveyers shown in figure 4. The conveyance velocity
was measured for different
(14 x 7 x 0.5 mm3 silicon objects) as shown in figure 8. The
speed was measured using a video recorder and counting the
number of frames (20 ms per frame). The moved distance was
determined by a comparison with the well specified reference
distance between the legs on the conveyer chip. The maximum
measured conveyance speed was 12 mm/s with an applied
voltage of 23 V (approx. 1.3 W) and 250 Hz. During the load
test when up to six silicon objects (weights) were stacked on
numbers of loads
each other the speed was reduced by using a lower heating
power. The velocity is load-independent (for loads less than
700 mg) as can be seen in figure 9. The difference between
theory and experiment is probably due to divergence in the leg
position between different legs (some are more erected than
others). The largest moved object was a 2000 mg weight as
shown in figure 10.
Position accuracy and roughness experiment
Besides high speed and load performance a micro-conveyer
system also needs to have good position accuracy. Accurate
positioning requires high precision actuation. Theoretically each
individual polyimide joint actuator can make extremely small
displacements. Then the accuracy only depends on the surface
roughness of the moved object because the vertical displacement
∆y of each leg has to be larger than the surface roughness of the
moved object to obtain the motion. The moved objects used in
our experiment are silicon dies with surface roughnesses of
approximately 2x10-3 µm (polished) and 2.5 µm (unpolished).
Both these movers show the same behavior concerning the
minimum required displacement of the leg despite different
surface roughnesses. At high frequencies and low power
dissipation which give a stroke length, ∆x less than 5-15 µm
and a vertical displacement ∆y between 2 and 7 µm (much larger
than the surface roughness of the moved object) we were not
able to obtain any conveyance motion at all. The object just
moves up and down. The reason for this is that we have
problems with small divergences of the static bending angle of
Number of load cycles
Life-time measurements with a three V-groove
polyimide joint. All bending angle data were measured at 2.5 Hz.
The standard uncertainty, uc in the measurements is 0.2°. It is
mainly caused by variation in the surrounding illumination.
Bending angle, α [°]
1 week (170 hours) @ 250 Hz
3 weeks (500 hours) @ 2.5 Hz
Bending angle (AC-on)
Bending angle (AC-off)
Silicon object (14X7X0.5 mm3)
Figure 8. Photo showing the conveyer moving a silicon
object (14 mm x 7 mm x 500 µm and 115 mg).
The maximum measured velocity is 12 mm/s (at 250 Hz and
216 mW/leg) using legs with a length of 500 µm and
four V-groove polyimide joints.
= 90 mW/leg
115 mg Si-object
700 mg Si-object
Velocity, v [mm/s]
Frequency, f [Hz]
Figure 9. Velocity, v, as a function of frequency, f. The
theoretical curve has been obtained by multiplying the
frequency by twice the measured horizontal displacement, ∆x
from figure 6.
Figure 10. The micro-conveyer during a load test. The
2 g weight shown in the photo is equivalent to 350 mg
on each leg or 16.000 times the weight of the legs shown
in figure 4 (b).