Content uploaded by Fred K. Forster
Author content
All content in this area was uploaded by Fred K. Forster on Aug 21, 2014
Content may be subject to copyright.
MEMS-Vol. 1, Microelectromechanical Systems (MEMS) — 1999
ASME 1999
TRANSPORT OF PARTICLE-LADEN FLUIDS THROUGH FIXED-VALVE
MICROPUMPS
Ling-Sheng Jang, Christopher J. Morris, Nigel R. Sharma, Ron L. Bardell and Fred K. Forster
Department of Mechanical Engineering
Campus Box 352600
University of Washington
Seattle, Washington 98195-2600
ABSTRACT
Micropumps designed for the flow-rate range of 100–
1000µl
min have been developed by a number of research
groups. However, little data is available regarding the ability
of various designs to directly transport liquids containing par-
ticles such as cells, microspheres utilized for bead chemistry,
or contaminants. In this study the ability of pumps with no-
moving-parts valves (NMPV) to transport particles was investi-
gated. The results showed that a NMPV micropump was able to
directly pump suspensions of polystyrene microspheresfrom 3.1
to 20.3µmin diameter. The pump functionedwithout clogging at
microspherenumberdensities as highas 9000particles
µlofsus-
pension, which correspondedto over 90,000 particles per second
passing through the pump at a flow rate of 600µl
min. Perfor-
mancewith polystyrenemicrosphereswas the sameas purewater
up to the point of cavitation. Microspheres manufactured with
negative surface charge cavitated less readily that other micro-
spheres studied that were manufactured without surface charge.
However, cavitation did not appear to be a function of micro-
sphere size, total surface area or number density. Thus pumping
polystyrene microspheres was found to be more affected by sur-
face effects than by size, surface area or number density within
the range ofparametersconsidered. In the case of chargedmicro-
spheres, the maximum flow rate was reduced by 30% compared
to pure water whereas for unchargedmicrospheres the maximum
flow rate was reduced by approximately 80%.
Address correspondence to this author at the above address or to
forster@u.washington.edu.
NOMENCLATURE
C fluid capacitance (m
3
Pa)
V Volume (m
3
)
NC Identifies suspensions of microspheres manufactured with
no surface charge
WC Identifies suspensions of microspheres manufactured with
surface charge
k Gas constant, 1 for isothermal process, ratio of specific heats
for adiabatic process
p absolute pressure (Pa)
β Bulk modulus (N
m
2
)
µ absolute viscosity (Pas)
ρ mass density (kg
m
3
)
φ Volume fraction of particles in suspension
c
pump chamber property
g
gas property
l
liquid property
p
particle property
s
suspension property
INTRODUCTION
Severaldesignsfor micropumpshavebeendevelopedbydif-
ferent groups based on various principles of actuation and vari-
ous types of valves. Most pumps are designed to work with pure
fluids, and in most cases there is considerable risk of valve clog-
ging if fluid with particles is directly pumped. Micropumps with
no-moving-parts valves (NMPV) are of interest because of their
503 Copyright by ASME
simplicity, reliability, ease of manufacture, and potential for be-
ing ableto pass particles (Forster et al., 1995; Olsson et al., 1995;
Bardell et al., 1997). However, there have been no studies on the
effectofdirectlypumpingparticles throughpumps of thisdesign.
Even without complications of particles interfering with the
action of valves that open and close, which are minimized in
NMPV pumps,particles may affectpumpperformanceindirectly
through changes in macroscopic fluid behavior caused by parti-
cles. Based on current pump models (Bardell et al., 1997) the
fluid properties that affect pump performance are mass density,
viscosity and compressibility. In terms of fluidic model parame-
ters these properties are associated with inertance, resistance and
capacitance.
Mass density affects fluid inertance, which for a conduit is
proportional to the product of density and length and inversely
proportional to cross-sectional area. Since the cross-sectional
area of the valves is relatively small, valve inertance has signif-
icant influence on pump operation. The density of a suspension
relative to the liquid phase is given by
ρ
s
ρ
l
1 ρ
p
ρ
l
1 φ (1)
where φ is the volume fraction of particles. For increasing parti-
cle mass density, significant relative motions between particles
and fluid may develop in NMPV pumps. This behavior may
affect the manner in which particles and fluid travel through
multi-channel fixed-geometry valves and could alter valve per-
formance.
Fluid resistance in a conduit is proportional to viscosity and
a strong function of the minimum transverse dimension. For ex-
ample, for a conduit of circular cross-section the fluid resistance
is inverselyproportionalto the forth power of diameter. Since the
transverse dimension of pump valves is relatively small, valve
resistance has a significant influence on pump operation. The
viscosity of a dilute suspension characterized by the Einstein re-
lation relates the effective absolute viscosity to that of the liquid
and to the volume fraction of particles
µ
s
µ
l
1 2 5φ (2)
(Probstein, 1989). When the size of particles in a suspension
approaches the dimensions of the pump valves, additional vis-
cous effects preclude modeling valve fluidic resistance based on
a simple macroscopic change in viscosity. In that case two phase
flow and lubrication theory may be required to accurately model
valve losses due to viscous effects.
The NMPV micropump operates near resonance by design,
and the resonant frequency is strongly dependent on capacitance
of the fluid in the pump chamber. At frequencies above reso-
nance a significant amount of membrane displacement that ide-
ally would result in outlet flow through the valves is consumed
by the capacitance of fluid in the pump chamber (Bardell et al.,
1997). Capacitance arises from fluid bulk modulus, which for a
suspension is given by
β
s
β
l
1 1 1 β
l
β
p
φ (3)
The capacitance of a suspension and gas in the pump chamber is
given by
C
c
V
s
β
s
V
g
kp (4)
whereV
c
V
s
V
g
, p is absolute pressure and k is a gas constant
that is unity for an isothermal process and the ratio of specific
heats for an adiabatic process (Rowell and Wormley, 1997). The
capacitance due to gas in the pumped fluid can have a dramatic
effect on pump behavior, i.e. the second term in the above equa-
tion can be much larger than the first. In addition to the basic
manner in which entrained gas affects pump chamber capaci-
tance, the role of particles in entraining such gas may signifi-
cantly contribute to changes in pump operation.
Due to the many factors described above that have poten-
tial effects on pumping suspensions, a series of experiments was
conductedtodevelop a morequantitativeunderstandingoftheef-
fects of pumpingparticle-laden fluids. The experiments included
means to investigate many of the factors associated with suspen-
sions. Steady flow tests, frequency analysis and basic pump per-
formance tests were used along with results of an existing sys-
tem model to investigate the individual factors associated with
suspensions.
METHODS
The methods usedin this studycan bedividedinto a descrip-
tion of the particular pump utilized, the preparation of the parti-
cle suspensions, the test for non-aggregation, steady flow tests,
frequency response tests and pump performancemeasurements.
Pump Description
The design of fixed-valvepump used for all tests is shown in
Fig. 1. The pump chamber and valves are shaded darkest. They
were etched in silicon to the same depth. An anodically bonded
rectangular Pyrex layer was used to seal the valves and chamber.
It also acted as a deformable plate where it spanned the pump
chamber. Shown shaded lightest and having the smallest diame-
ter is the piezoelectric element. It was bonded to the Pyrex with
conductive epoxy to form electrical contact with the bottom sur-
face of the piezoelectric element. The epoxy is shown shaded
darker than the smaller piezoelectric element and includes the
electric connection region on the upper right side of the pump.
The valves connected to plenum chambers where connections to
504 Copyright by ASME
Figure 1. CONFIGURATION OF A T45/4X26-MM DIAMETER
PUMP
.CHAMBER AND VALVES ARE SHOWN DARKEST, PIEZO-
ELECTRIC ELEMENT IS SHOWN LIGHTEST AND CONDUCTIVE
EPOXY IS SHOWN IN AN INTERMEDIATE SHADE OF GRAY
.THE
DIRECTION OF NET FLOW IS INDICATED WITH AN ARROW
.
an inlet reservoir and the load were made. Two different mi-
cropumps were used. They only differed in the etch depth of the
chamber and valves. Deep reactive ion etching (DRIE) was used
to etch one pump to a depth of 117µm and the other to a depth of
156µm. The pump chamber diameter was 6mm, and the cover
plate was 500µm thick. The piezoelectric material used for the
driving element was PZT-5A (PSI-5A-S2, Piezo Systems, Inc.,
Cambridge, Massachusetts). It was 5mmin diameter and 190µm
thick. The transverse dimension of the valves was approximately
114µm. Further details on the type of fixed-valve pump used
have been published previously (Bardell et al., 1997).
Sample Preparation
The suspensions used consisted of two different types of
polystyrenemicrospheres. One typehad a nominaldiameter(and
standard deviation) of 20.3(0.33)µm (PS07N, Bangs Laborato-
ries, Fishers, Indiana) with a specific gravity of 1.062. These mi-
crospheres had no special surface treatment and in this study are
referred to as NC for not charged. Another type of microsphere
used was obtained in three different diameters (and standard de-
viation) of 3.1(0.11),4.6(0.19)and 7.9(0.85)µm, respectively (1-
3000, 1-4500, 1-8000, Interfacial Dynamics, Corp., Portland,
Oregon) with a specific gravity of 1.055. These microspheres
were negatively charged with sulfate functional groups on the
surface and had a charge density between 5.4 and 5.6µC
cm
2
.
In this study these microspheres are referred to as WC for with
charge.
To prepare suspensions of different microsphere concentra-
tion, a known volume of de-ionized water filtered at 0.2 um was
combined with a predetermined volume of stock microsphere
suspension based on the manufacturers’ specification of percent
solids. Precision micropipettes (MW128, Chemglass, Vineland,
New Jersey) were used to measure the volumes. The diluted WC
microsphere suspension was degassed for 10 min with a vacuum
pump (N810.3FT8, KNF Neuberger, Inc., Trenton, New Jersey)
while the sample was agitated in an ultrasonic bath (Bransonic
12, Branson Instr. Co., Shelton, Connecticut) and kept at room
temperature. In the case of the NC microspheres, de-ionized wa-
ter was degassed as described above, after which the predeter-
mined amount of stock suspension was added. The latter pro-
tocol ensured the resulting diluted suspension consisted of non-
aggregated microspheres.
The diluted suspensions were introduced into pumps in a
manner that completely displaced all air. This was accomplished
by first filling the pumps with ethanol.
Aggregation Test
Particle suspensions were checked for aggregation as a part
of all pump tests. Before and after each test a small amount of
the fluid sample was collected with a syringe labeled for that
particle size. An inverted microscope (IM35, Zeiss, Germany)
was used in conjunction with a high performance CCD camera
(4912-2000/0000, Cohu, Inc. San Diego, California) whose im-
age was displayed on a video monitor. A drop of suspension
was placed on a glass slide and viewed through the microscope.
A piece of metal foil was used to reflect forward-projected light
backthroughthemicrospheresto increasetheirvisibility. If more
than approximatelyfive percent of the particles in a field of view
were aggregated in groups of three or more, the suspension was
determined to be aggregated and it was discarded. Otherwise the
suspension was determined to be non-aggregated, and the test
data was used.
Steady Flow Resistance
A syringe pump (Model 200, KD Scientific, Boston, Mas-
sachusetts) was usedto providea known flow rate for steady flow
resistance tests on the pumps. A precision syringe (GASTIGHT,
Hamilton Company, Nevada) was utilized in the syringe pump.
It was connected to one branch of a plastic Y-section from an in-
travenous filter set (SFE-2017SL, B. Braun Medical Inc., Beth-
lehem, Pennsylvania) using standard Luer-lock fittings. A pres-
sure transducer (EPI-127, Entran Sensors & Electronics, Fair-
field, New Jersey) was connected to the other branch, and the
outlet was connected to the pump. The pressure transducer was
previously calibrated against a mercury manometer. Flow rates
of 1000 to 5500µl
min were applied, and the DC pressure trans-
ducer output was read on an oscilloscope.
Frequency Response Tests
Frequency response of the pump was measured in terms of
the peak velocity of the center of the driving piezoelectric ac-
505 Copyright by ASME
Tabl e 1 . CASE DEFINITIONS FOR ALL THE POPULATIONS OF MICROSPHERES USED IN THIS STUDY.WCREFERS TO THE MICRO-
SPHERES MANUFACTURED WITH ADDED NEGATIVE SURFACE CHARGE, AND NC REFERS TO THE MICROSPHERES MANUFACTURED
WITHOUT ADDED SURFACE CHARGE
. β
p
WAS TAKEN TO BE 5 10
9
N m
2
.
case type size concentration number density surface area ρ
s
ρ
l
1 µ
s
µ
f
1 β
s
β
l
1
(µm) (µg ml) (beads ml) 10
6
(µm
2
ml) 10
6
10
6
10
6
10
6
A WC 3.1 70 4.19 126 4 166 37
B WC 3.1 150 9.22 278 8 356 80
C WC 4.6 230 4.19 278 12 545 123
D WC 7.9 390 1.42 278 20 924 208
E WC 7.9 1100 4.19 821 57 2608 587
F NC 20.3 250 0.0537 70 15 598 132
G NC 20.3 1000 0.215 278 58 2354 529
tuator, as a function of frequency at a low driving voltage, and
with the pumping system filled with suspension. A vibrometer
(Polytec OFV 2600, Germany) was used to measure the veloc-
ity. A Sweep/Function Generator (Model 19, Wavetek, United
Kingdom) and a piezo amplifier (EPA-102, Piezo system Inc.,
Cambridge, Massachusetts) were used to drive the piezoelectric
actuator with a sinusoidal signal. Twenty-cm long silastic tubes
having a diameter of 1mm (62999-166, VWR Scientific, Bris-
bane, California) were connected to the inlet and outlet ports of
the pump. The inlet and outlet tubes were inserted into open
reservoirs, the liquid surfaces of each were at the height of the
pump.
Pump Performance Tests
Pump performance tests consisted of measuring the zero
load output flow rate and the blocked flow outlet pressure head.
For flow measurement the inlet and outlet tubes were configured
as described above. A micro “bucket and stop watch” method
was utilized. The outlet reservoir was situated on an electronic
scale (1205MP, Sartorius, Westbury, N.Y.), andchange in weight
over a know time interval was used to calculate flow rate. The
blockedflowpressure was measuredwith the pressure transducer
that was used forthe steady flow tests connected to the end of the
outlet tube.
RESULTS AND DISCUSSION
Seven different suspensions were considered from four dif-
ferent sized microspheres. Table 1 summarizes the factors that
differentiated each suspension. In the table values for relative
density, viscosity and bulk modulus were calculated with Eqs. 1,
2 and 3. Four cases were at concentrations that yielded the same
surface area but with distinct sizes, and three cases were at con-
centrations that yielded the same number density but also with
distinct sizes. Cases of equal surface area and number density
were chosen to highlight any effects that might correlate with
either parameter.
To perform all tests each suspension was pumped for a min-
imum of 30 minutes. During no test was any clogging of a pump
observed. For a given number density and flow rate, the actual
number of particles successfully pumped per unit time was sub-
stantial. For example, when suspension B was pumped at a flow
rate of 600µl
min, a pumping rate of 5.4 million particles per
minute was attained.
Examples of suspensions with and without aggregation are
shown in Figs. 2 and 3. From such micrographs, it was easy
to assess whether a suspension aggregated. Aggregation was a
problem with the NC microsphere suspensions, which were de-
gassed using a modified procedure due to their tendency to ag-
gregate. The NC microsphere suspensions eventually aggregated
over time after preparation, but with the method used to monitor
aggregation,pump data waseasy to categorizeinthatrespect. All
data reported below correspondedto suspensions that did not ag-
gregate and therefore corresponded to well-defined microsphere
sizes.
The purpose of measuring pressure drop across a non-
operating pump as function of steady flow rate was to investi-
gate the role of suspension parameters on viscous effects asso-
ciated with the NMPV pump. Since the flow was steady, it was
assumed that any effects of inertance or capacitance were negli-
gible compared to the case of oscillating flow, i.e. during pump
operation. The pressure drop for pure water and various suspen-
sions is shown in Fig. 4 as a function of volume flow rate in both
the forwardand reverse directions. The resistance is given by the
506 Copyright by ASME
Figure 2. AGGREGATED 20.3µmNCMICROSPHERES IN A
DROPLET OF SUSPENSION
,CASE G IN TABLE 1, RECORDED
WITH A
CCD CAMERA AND A 4X OBJECTIVE LENS.
Figure 3. NON-AGGREGATED 7.9µmWCMICROSPHERES IN
A DROPLET OF SUSPENSION
,CASE D IN TABLE 1, RECORDED
WITH A
CCD CAMERA AND A 10X OBJECTIVE LENS.
local slope of the pressure versus flow curve. There was very lit-
tle difference between the resistance of pure water, and the resis-
tance of any of the particle suspensions. From the data presented
it can be concluded that viscous effectsdue to the particle param-
eters were negligible. Since Eq. 2 predicts a very small effect on
the effective viscosity of any of the suspensions utilized, the data
supports that prediction. Less obvious, however, is that even in
the case where the valve minimum dimension to particle size ra-
0 1000 2000 3000 4000 5000 6000
0
1
2
3
4
5
6
7
8
Forward data
Reverse data
flow (µl/min)
pressure (m H
2
O)
water
B
C
D
E
F
G
Figure 4. PRESSURE DROP VERSUS STEADY FLOW RATE FOR
PURE WATER AND VARIOUS PARTICLE SUSPENSIONS DE
-
SCRIBED IN TABLE 1 AND THROUGH THE PUMP IN FIG.1.
tio was approximately six, no detectable effects were seen in the
resistance values in forward or reverse flow. Thus, the current
NMPV pump is not influenced by viscous flow effects caused
by particle sizes as large as 20µm for the concentrations consid-
ered. This is very encouraging because particles of that size are
larger than the valve clearance of many micropumps designed
with moving valve geometry.
The pump used for the above steady-flow tests was 25%
shallowerthan the pumpused for all the othertests. Since nodif-
ferencein flowresistance betweenanysuspensionandpure water
was detected for that pump, no difference would be expected for
the deeper pump used for the rest of the tests. It should also be
noted that for all steady flow resistance results, reverse flow data
exhibited higher resistance than the forward flow data. This was
expected, because this difference allows the valves to generate
a net flow in the direction of smaller resistance when subjected
to an oscillating pressure. This characteristic of fixed-geometry
valves allows them to function as pump valves.
With the knowledge from the steady-flow tests that viscous
effects on pumping are unchanged by any of the parameters of
the suspensions considered, the dynamic response of the pump
was used to investigate the effect of microspheres on fluid capac-
itance. Figure 5 is a plot of the frequency dependent behavior of
the pump around the resonant frequency associated with pump-
ing, i.e. a resonance that is related to fluid flow in the valves
and dependent on the fluid resistance and inertance in the valves
and the capacitance of fluid in the pump chamber (Bardell et al.,
1997). The response shown in the figure is the centerline veloc-
ity of the membrane. Other output parameters such as valve flow
507 Copyright by ASME
3 3.5 4 4.5 5 5.5 6
1000
2000
3000
4000
5000
6000
7000
8000
9000
peak−to−peak velocity ( µm/s)
frequency (kHz)
water
A
B
C
D
E
F
G
Figure 5. MEMBRANE PEAK CENTERLINE VELOCITY VERSUS
FREQUENCY AT
15V PEAK-TO-PEAK FOR PURE WATER AND
VARIOUS PARTICLE SUSPENSIONS DESCRIBED IN
TABLE 1.
3 3.5 4 4.5 5 5.5 6
25
30
35
40
45
50
55
60
frequency (kHz)
membrane velocity gain (dB)
before
after 5% increase
in capacitance
Figure 6. CALCULATED EFFECT OF CHAMBER CAPACITANCE
ON THE FREQUENCY RESPONSE OF PUMP MEMBRANE CEN
-
TERLINE VELOCITY.THE INPUT PARAMETERS CORRESPOND
TO THE PUMP SHOWN IN
FIG.1.ZERO dB 1µm s AT A 1V
PEAK-TO-PEAK DRIVING VOLTAGE.
and chamber pressure would show similar behavior. It is clear
that there is no discernable difference in magnitude or frequency
of the peak response. However, the results of a linear model
(Bardell et al., 1997) of the pump in Fig. 6, show that with just a
five percent change in the amount of entrained gas in the cham-
2.5 3 3.5 4
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
outlet pressure (m H
2
0)
frequency (kHz)
water
A
B
C
D
E
F
G
Figure 7. BLOCK LOAD OUTLET PRESSURE AS A FUNCTION
OF FREQUENCY AND A
170 V PEAK-TO-PEAK DRIVING SIG-
NAL FOR PURE WATER AND VARIOUS PARTICLE SUSPENSIONS
DESCRIBED IN
TABLE 1.
ber, the resonant frequency shifts noticeably. It is possible that a
similar variation in valve inertance could cause a corresponding
change in resonant frequency, but the changes in density due to
particles shown in Table 1 are less than 0.01%. Thus the capac-
itance effects on pump operation were unchanged by any of the
parameters of the suspensions considered.
Having found that inertance, resistance and capacitance ef-
fects of the various suspensions did not cause any observable
effects on pump characteristics, one would expect that the pump
performancewould be unaffected by any of the suspensions con-
sidered compared to water. Figure 7 shows that to be the case
in terms of the block load pressure versus frequency for a typ-
ical driving signal of 170V peak-to-peak up to a frequency of
approximately 3000Hz, after which both NC suspensions cavi-
tated. To look more closely at any possible trend in performance
for a frequency at which all suspensions were pumped data from
Fig. 7 at 2.9kHz and similar data for no-load flow rate were uti-
lized to generate the pump performance curve shown in Fig. 8.
The lines drawn between data points are typical of performance
obtained with more data (Bardell et al., 1997). However, the fig-
ure shows no correlation between performance and microsphere
size, surface area or number density.
A very different result was obtained when the maximum
performance attainable with each suspension was compared as
shown in Fig. 9. The maximum performance was determined
with an input signal of 170V peak-to-peakand by increasing the
frequency of the input signal until cavitation occurred. Clearly
the results grouped into three distinct regions corresponding to
508 Copyright by ASME
0 50 100 150 200
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
flow rate (µl/min)
head (m H
2
O)
water
A
B
C
D
E
F
G
Figure 8. PUMP PERFORMANCE CURVES AT A TYPICAL OPER-
ATING POINT OF 2900 HZ AND 170 V PEAK-TO-PEAK DRIV-
ING SIGNAL, FOR PURE WATER AND VARIOUS PARTICLE SUS-
PENSIONS DESCRIBED IN TABLE 1. LINES DRAWN FOR CLAR-
ITY.
pure water (having the highest performance), then the WC sus-
pensions and finally the NC suspensions. This grouping is also
evident in Fig. 7 where the maximum NC frequency is approx-
imately 3kHz, where the maximum WC frequency is 3.5kHz
and where the maximum frequency before cavitation occurred
for pure water is 3.7kHz. Thus, the primary parameter that pro-
duced variations in pumping characteristics was the tendency to
cavitate. Both WC and NC microspheres were polystyrene and
hydrophobic according to the manufacturers. However the WC
negatively charged particles were significantly more resistant to
cavitation. We conclude that the tendency for particle suspen-
sions to cavitate while being pumped is significantly affected by
surface chemistry. This effect was significant enough to over-
ride a seven-fold difference in microsphere diameter, an 80-fold
difference in number density and a 12-fold difference in surface
area.
ACKNOWLEDGMENT
This work was partially supported by DARPA/ ETO, con-
tracts N660001-97-C-8632 and F30602-98-2-0151. The authors
also acknowledge support from the Stanford Nano Fabrication
Facility for microfabrication including DRIE.
References
Bardell, R. L., Sharma, N. R., Forster, F. K., Afromowitz, M. A.,
and Penney, R. J. (1997). Designing high-performance
0 200 400 600 800 1000
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
flow rate (µl/min)
head (m H
2
O)
water
A
B
C
D
E
F
G
Figure 9. MAXIMUM PUMP PERFORMANCE ATTAINED BE-
FORE CAVITATION WITH 170 V PEAK-TO-PEAK DRIVING SIG-
NAL FOR PURE WATER VARIOUS AND PARTICLE SUSPENSIONS
DESCRIBED IN
TABLE 1. EACH CASE CORRESPONDED TO THE
DRIVING SIGNAL FREQUENCY THAT YIELDED THE HIGHEST
PERFORMANCE
.LINES DRAWN FOR CLARITY.
micro-pumps based on no-moving-parts valves. In
Microelectromechanical Systems (MEMS), volume DSC-
Vol.62/HTD-Vol.354,pages 47–53. ASME IMECE.
Forster, F. K., Bardell, R. L., Afromowitz, M. A., Sharma, N. R.,
andBlanchard, A. (1995). Design, fabrication andtesting of
fixed-valve micro-pumps. In Proceedings of the ASME Flu-
ids EngineeringDivision, volume 234, pages 39–54. ASME
IMECE.
Olsson, A., Stemme, G., and Stemme, E. (1995). A valve-less
planar fluid pump with two pump chambers. Sensors and
Actuators A, A47(1-3):549–56.
Probstein, R. F. (1989). Physicochemical Hydrodynamics: an
Introduction. Butterworths, Boston.
Rowell, D. and Wormley, D. N. (1997). System Dynamics an
Introduction. Prentice-Hall, Inc.
509 Copyright by ASME
