@
The
experiences in the development of active microvalve using PZT
generated standing wave
Z.
Yang1,2 and R. Maeda1*
1Mechanical Engineering Systems, AIST, Tsukuba, Japan 305-8564
2
NEDO, Tokyo, Japan
ABSTRACT
The design, fabrication and testing efforts for development of active microvalve
were presented.
Microfabrication technologies were used in these efforts. The valve was designed
to be actuated using PZT generated standing wave. It has not worked yet. The
problems on actuating principle, structures, materials and fabrication processes
were shown here. We like to share these experiences with fluidic MEMS developers
to avoid the similar errors we have already failed.
INTRODUCTION
Microvalve
is one of the elemental devices in fluidic MEMS. With
the development of micromachining technology, the methods
for creating networks of microchannels are well established for silicon, glass
and polymeric substrates. PZT-actuated micropumps
and micromixers have achieved high performance [1, 2].
Various
microsensors have been reported. For microchannels, there are
no movement parts. Their fabrications rely on sophisticated etching techniques
and bonding techniques which have been well-developed. The successes of PZT-actuated
micropumps and micromixers rely on using the high frequency characteristics of
PZT to compensate its relatively small displacement. In this way, PZT elements
convert the electrical energy into mechanical energy effectively, which
performed as pump or mixer. However,
the development of microvalve has its own difficulties. The fundamental problem
for microvalve is to control the leakage over a movement part. There are several
active microvalves have been reported. Some of them are already commercially
available. However, further improvements
are still required.
The
working principle and design of our microvalve
Our
approach is to extend the concept of micropumps and micromixers. Large movement
for opening/closing operation of valve could be achieved using high frequency.
Ultrasonic standing wave was employed for microvalvefs operation. One property
of standing wave is to move particles to its Ι/2 (Ι: wavelength) wave node.
The Ι/2 of 1 MHz wave in water is 750 Κm. By adjusting the phase of standing
wave, particles can be moved hundreds of Κm in water, which is very difficult
to reach using the PZT direct actuation. We try to use this movement for
operating microvalve.
A
silicon/glass anodically bonded structure was chosen for flow channels (Fig. 1).
The piston for opening/closing the valve was a polyimide disc in the diameter of
100 Κm. The channel depth was determined by the precision of thickness of the
polyimide disc. The states of opening and closing of the valve were shown in
Fig. 2, respectively. Ultrasonic standing waves were introduced into the fluidic
channel through a pair of horns forming on silicon by actuating PZT ceramics
(see Fabrication part). The position of the piston was designed to be controlled
by changing the phase of standing wave from the pair of horns.
The length of the horns is defined by the Ι/2 in silicon. The Ι/2
in silicon at 1 MHz is 4.2 mm. The length of horn
is 8 mm (Ι/2 of 525 kHz) for
this experiment.
The
design was focused to minimize the escape of ultrasound to other part of the
device but focus into fluidic chamber. From the front of view in Fig. 1., the
anodically bonded chip frame and the wall of flow channel look darker then the
area of the pair of horns does.
Fabrication
Conventional photolithographic methods were used in
the fabrication of the device.
It was a 4 masks processes (one for polyimide piston, 1 for glass and 2 for the
front and back sides of Si, respectively).
Fig.
2. The pistone of a polyimide disc in the states of valve opening (left) and
closing (right). The diameter of the polyimide disc is 100 Κm. Most parts of
silicon were etched away to focus the ultrasound inside the chamber.
1
The fabrication of Σ100 Κm polyimide disk for the piston of the valve
A thermally oxidized 100 mm
silicon wafer with 2 Κm SiO2 was used as substrate. The SiO2 was
used as a scarified layer. A negative type photosensitive polyimide (PI 2729, HD
Microsystems) was spinning-coated (1H-DX2,
Mikasa, Japan) on the substrate at 350 rpm/7 s + 800 rpm/15 s + 1400
rpm/30 s to get a thick polyimide membrane. Prebaking was processed under 85 oC/4 min + 105 oC/4 min + 125 oC/4 min. Exposure was for 480 s using Unionfs mask aligner. Post- exposure
baking was proceeded at 85 oC/1 min and then the
wafer was developed (D6180, HD Microsystems) for 5 min and rinsed (9180, HD
Microsystems) for 2 min. The postbaking was proceeded at 150 oC/3 min + 200 oC/3 min. Then the wafer
was moved to an oven purged with N2. Temperature was gradually increased from
250 oC to 300 oC in 50 min and was kept at 315 oC for 60 min. The thickness of polyimide disks were shown in Fig. 3.
A protecting tape (ReverAlpha) was stuck on the substrate of the polyimide
side and then immersed it into buffered HF solution (LL6) for 1 week to etch
away SiO2 sacrificed layer. Then the polyimide disks were on the tape.
Fig. 3. The profiles from
Dektak3 of polyimide disks cured after 200 oC
(upper line) and 315 oC (lower line), respectively. After the finial
thermal treatment, the edges of the disks deformed and their thickness shrunk
10 %. The difference on diameters is not true here. These differences in
diameters were just cased by the measuring line missed the diameter line.
2
Silicon and glass fabrication processes
The front side of
a 100 mm silicon wafer was etched into 24 Κm at front side (for flow channel) and 320 Κm at backside (to isolate ultrasound from other part of the device).
DRIE from STS technologies was used for the etching. A
9-Κm-thick (postbaking at 150 oC for 30 min) photoresist layer (S
1830, Shipley, USA) was used as the etching mask. The thick photoresist layer
was spin-coated onto silicon at 3000 rpm for 1 s.
The patterns on
100 mm glass wafer (Pyrex glass #7740,
Iwaki Glass, Japan) were formed by HF isotropic etching (50% HF-69% HNO3-H2O in
volume ratio
of 2:1:2) at room temperature. A 200-Å-thick chromium layer, topped by a
1500-Å-thick gold layer, was evaporated as the etching mask.
Before anodically bonding,
both of the etched silicon and glass wafer were treated with diluted HF solution
(1:80 in volume with water) for 20 s. A polyimide disk was manually put into
each valve chamber area on silicon using a tweezers precision. A dummy glass
wafer with sputtered Pt/Ti layers were inserted between glass wafer and cathode.
This prevented the damages on glass at contacting area around cathode. Anodic
bonding was proceeded at 370 oC and then 700 V was added for 15 min.
After a Pt/Ti background electrode was sputtered onto
the Si side, individual chips were diced
using diamond dicing saw (MD-60, Okamoto). Then, bulk piezoelectric PZT ceramics
(1 mm x 2 mm x 0.1 mm) were adhered directly onto the sidewalls of the silicon
horns. Finally, the each die was adhered on a DIP IC socket. Wires for electric
power supply and silicone tubes for fluid were connected to the device,
respectively. The assembly processes was illustrated in Fig. 4.
Fig. 4. Illustration of the
assembling processes (not to scale). After silicon was patterned (backside was
not shown), the polyimide pistons (only one, in the middle of silicon, was
shown here) were put into the flow chambers and then covered with a patterned
glass. These two wafers were anodically bonded. After each chip was separated
by dicing saw, chip was adhered on a DIP frame and then 2 pieces of PZT
ceramics were mounted on the sidewall of the pair of horns, respectively. The
whole assembling processes were finished after wiring the electrodes of PZTs
and connecting 2 silicone tubes (not shown here) to the inlet and outlet of
flow chamber, respectively.
Experimental
Water was used to evaluate the performance of the
device. External pressure was applied to water using a fluid dispenser
(Custom-ordered from Musashi Engineering Inc., Japan) to maintain a continuous
flow. A
2-channel function generator (Sony Tektronix AFG320,
Japan) connected to two power amplifiers
(NF Electronic Instruments 4010, Japan) was used to generate the square wave (525
kHz) for PZTs excitation. The phase of the two driving
waves from the function generator was tunable.
Results and discussion
The design worked
but there were serious drawbacks. The piston can be operated only one time. When
the piston touches an edge of silicon, it will never move anymore.
During the
fabrication processes, the polyimide disks shrunk about 10% of their thickness
after high temperature treatment. The edges of the disk were deformed but would
not affect diskfs movement (Fig. 3). The design of the horn is not optimized.
The length of the horns was determined by the half wavelength of 525 kHz in
silicon but the width at the end of horn is quite small than wavelength. The
transferring loss of ultrasound in horn is unknown. The isolation of the horns
from the other parts of the device was concerned carefully. Both silicon and
glass wafers were etched away around the horns.
We observed that
the piston was not moved followed the flow of water or air under the external
pressure upper to 0.2 MPa, which is the maximum pressure for the connection
using silicone tube.
When
standing wave was added, the piston did not move immediately. By tuning the
phase of the standing wave, the piston was observed to move following the water
flow.
After the piston
touched the silicon wall, it did not response to the changes in phase, frequency
and/or voltage of the actuating wave but stopped there. The principle of using
standing wave to operate particles works only for particles between the pair of
horns. When polyimide disk attached to the silicon wall, it became one part of
horn. Because the impedance between silicon and polyimide is lower than that
between silicon and water. Thus, when polyimide disk attached to silicon wall,
the ultrasound would transmit through polyimide and then emitted into water. The
standing wave is forming in water and will not affect to the polyimide disk any
more.
This device also
failed to work even as a single use valve. When the polyimide disk was at the
close position, there was no significant change in flow rate. The leakage was
relatively large. The piston could not be opened using vacuum or introducing
water or air from the outlet of the device. The difficulties on geometric
control were mainly from etching of flow channel on silicon using DRIE. The
homogeneity of DRIE etching does not reach 10% for 100 mm wafer. Flow channels
were etched for 24 Κm in depth, it was in error by more than 2 Κm. Other factor for the leakage was the deformation of
polyimide disk, as shown in Fig. 3.
The
first challenge of using high frequency ultrasound to operate microvalve was
failed.
