CUN. CHEM.40/1,43-47(1994)
#{149}
Automation
and Analytical
Techniques
Manipulation and Flow of Biological Fluids in Straight Channels Micromachined
in Silicon
Peter Wilding,’
Joseph Pfahier,2 Haim H. Bau,3 Jay N. Zemel,4 and Larry J. Krlcka”5
of minute sample volumes is a major analytical
challenge that requires an understanding of fluid flow in
microstructures. Accordingly, flow dynamics of biological
fluids and cell suspensions in straight glass-capped silicon microchannels (40 to 150 m wide, 20 and 40 m
deep) were studied. We demonstrated that these microstructures are appropriate components
for microfluidic
analytical devices. Different fluids were easily manipulated in the microchannels, and measurements of flow
rate as a function of pressure for whole human blood,
serum, plasma, and cell suspensions revealed non-Newtonian behavior. By means of micromachined filters (5
m) located in channels, blood cells and microparticles
were effectively separated from nanoliter-sized samples,
clearly indicating the future role of microstructures for a
variety of analytical purposes.
Analysis
Indexing
Terms: viscosity/shear rate/non-Newtonianfluids/micro-
channels
laboratory
analysis
has been the
progressive reduction in the volume of sample required
for assay. Early methods required at least 1 mL of sample, but modern-day
techniques can analyze volumes as
small as 1 L (1). Various factors are responsible
for
this trend, including concerns over the hazard posed by
bloodborne
pathogens
in some clinical samples (human
immunodeficiency
virus; hepatitis
A, B, and C), patients’ convenience
(fingerstick samples), reduction in
reagent usage, and cost of testing. The next analytical
challenge is to further reduce sample volumes to the
nanoliter-picoliter
range (10_9_10_12
L). Many clinically important substances are present at micromole per
liter concentrations
(10-s mol/L); thus, 1 nL of sample
could contain 1 fmol of analyte,
which corresponds
to
100 miffion molecules. This number of molecules is well
within the detection range of several analytical methods
(e.g., chemiluminescence
and fluorescence)
(2). The major hurdle to implementing
analysis on nanoliter and
picoliter volumes is that the microfluidics required for
handling such small volumes has not been developed.
We believe that micromachined
sfficon structures incorporating interconnected
microchannels
and chambers will provide the necessary devices for handling and
analyzing
very small volumes of sample (3, 4). Silicon
micromachining has been used to produce micromeA trend
in clinical
Departments
of ‘Pathology and Laboratory Medicine, ‘Mechanical Engineering and Applied Mechanics, and4 Electrical Engineering, University of Pennsylvania, Philadelphia, PA 19104.
‘APD Cryogenics, Allentown, PA 18103.
5Author for correspondence. Fax 215-662-7529.
ReceivedJune 16, 1993; accepted September 10, 1993.
chanical devices such as valves, filters, cantilevers,
rotors, and gas chromatographs
(5-9). This technology is
readily
adaptable
to the development
of microfluidic
analytical
devices.
We used photolithographic
techniques to fabricate
straight
channels,
20-40 m deep,
and filter systems
of different
geometries
on silicon
chips. The advantage
of this new test system is that
disposable,
highly reproducible,
complex microchannel
networks
can be batch-manufactured
inexpensively.
We report here the flow characteristics
of biological
fluids and blood cell suspensions in straight channels of
differing size with volumes in the nanoliter range.
MaterIals and Methods
Fabrication of microchannels.
Microchannels and mlcrochannel
structures (Fig. IA) were fabricated by using
planar photolithography or reactive ion etching in silicon (400-sm-thick
wafers, crystallographic
orientation
<1,1,0>,
anisotropic
KOH etchant)
(SGS Thomson,
Montgomeryville,
PA) with photolithographic
masks
(Align-Rite, Santa Clara, CA). The wafers were diced
into 17 x 11 mm chips and then sealed with 1.58-mmthick Pyrex glass (Mooney Precision Glass, Huntington,
WV) by using a diffusive bonding technique (10). An
entrance and an exit port (500 x 500 m) were etched
through
the 400-pin-thick silicon at each end of the
11.7-mm-long channel. For each batch of channels, the
depth of one channel was measured with a surface proifiometer with a precision of ± 1%. Channels produced
with the anisotropic etchant had a trapezoidal crosssection (Fig. 1B).
Flow studies. We quantitated
fluid flow in microchannels with the experimental
apparatus
shown in Fig. 2. A
chip was held in a custom-built
holder (Faulkner
Instruments, Pitman,
NJ) on a microscope stage (Aristomet,
Wild Leitz, Heerbrugg,
Switzerland).
The inlet port of
the chip was connected to a syringe, housed in a syringe
driver (Model 350; Sage Instruments,
Boston, MA), and
actuated by a stepper motor (marimum
force 20 N). A
load cell measured
the applied force from which the
driving pressure was deduced. The force measurement
was corrected for frictional losses in the system, i.e., due
to friction between the syringe plunger and the syringe
barrel and pressure losses in the syringe itself. The force
required to overcome these frictional losses was determined by filling a syringe with distilled water and then
measuring
the force required to empty the syringe as a
function of flow rate. The functional losses ranged from
5% to 20% of the total force at the highest to the lowest
pump speed, respectively.
The flow rate for each of the
speeds was obtained by timing the syringe plunger displacement.
All measurements
were conducted
under a
CUNICAL CHEMISTRY, Vol.40, No. 1, 1994 43
A
LOADca
I
1=c
YRDRNER
Fig. 2. ExperImental
B
apparatus
for observing and recording flow In
silicon microchannels.
Flow Channel-\
VDU, video display unit; VCR, videocassette recorder.
Inlet Sump
Outlet Sump
SIllcon-i
A
A
a. Top View
Row Channel
and plain tubes. A leukocyte (WBC) suspension was
prepared by using Lymphocyte
Separation
Medium (Organon Teknika, Durham, NC) (11). Washed erytbrocytes (RBCs) were prepared with isotonic saline. Latex
microbeads (5.78 pin diameter) were purchased from
Polysciences
(Warrington, PA).
Protein-coated
channels.
The inside surfaces of some
of the microchanneis
were coated with albumin by filling them with a solution of albumin (50 g/L) (Sigma, St.
Louis, MO), and then drying the channels in a microwave oven.
Results
Inlet Sump
Silicon
-
Outlet Sump
b. Cross-Sectiona’View (Sect. A-A)
c. Channel Cross-Section(Sect. B-B)
Fig. 1. (A) Silicon chip with two straight channels (airows) (upper
60 ian wide, lower 80 cufl wide); (B) Schematic of silicon
channel
chip.
microscope to detect channel clogging. The pressure
losses in the syringe and supply piping was <1% of the
total pressure drop.
Flow in the channels was observed and recorded with
a black and white television camera (Dage-MTI, Michigan City, IN) and a videocassette recorder (PVM-122;
Sony, Teaneck, NJ). No failures were detected in the
diffusive bonded glass-silicon
junctions in any of the
chips studied. All studies were performed
at ambient
temperature
(22.3 ± 1.5#{176}C)
in an environmentally
con-
trolled laboratory.
Biological
samples
healthy
44
volunteer
and microbeads.
Blood from a
(J.P.) was collected into heparinized
CUNICALCHEMISTRY,Vol. 40, No. 1, 1994
We investigated
the flow of complex fluids (whole
blood, serum, washed RBCs and WBCs) from a 500 x
500 m entrance sump into straight channels (11.7 mm
long) with cross-sectional dimensions
80 &m wide (top)
x 20 m deep (volume -17 nL) and 150 m wide (top)
x 40 pin deep (volume -64 nL). The results of our
measurements
are depicted in Figs. 3 and 4 for the
40-pin- and 20-pin-deep
channels, respectively.
Figs. 3A
and 4A show the pressure head (the pressure needed to
drive the flow) as a function of flow rate. Figs. 3B and 4B
depict the pressure head needed to transport
various
biological fluids (APB) normalized
with the pressure required to transport water CAP) at the same flow rate
(relative viscosity &R = P3/iP).
Although it is conventional
in fluid mechanics topresent results as functions of the Reynolds number (crosssectional averaged velocity x effective diameter/viscosity), we have not done so because of possible ambiguity
in the definition of the viscosity of biological fluids that
exhibit non-Newtonian
behavior.
150-pm-wide
x 40-pin-deep channel. Flow rate as a
function of pressure head and relative viscosity for distilled water, serum, and RBCs are shown in Fig. 3. The
pressure required to achieve a given flow rate increased
as the complexity
and viscosity of the fluid increasede.g., for a flow rate of 0.25 mLlmin, distilled water required a pressure of-75 kPa, whereas the identical flow
rate with whole blood required
>250 kPa (Fig. 3A).
Whole blood,, diluted 3500:1 in isotonic saline to give the
same cell concentration
as the WBC preparation,
required less driving force than the WBCs at a given flow
rate. This is because whole blood is made up primarily
of
RBCs, which are highly flexible and much smaller than
A
1 600
o
Distilled Woter
#{149}Serum
A
RBC in saline
500
1 400
450
I
400
0
- 300
I
4)
250
(4
a-
200
-
-
-
1000
(4
4)
a-
800
600
400
150
100
-
-
p1200
-
350
a-
-
-
200
I
0
0.02 0.03 0.04 0.05 0.06 0.07
0.08 0.09
Flow Rote, mL/min
50
0
0.03
0.09
0.15
0.22
0.28
0.34
0.40
Flow Rote, mL/min
0.46
0.53
0.59
0.65
0.10
0.11
0.12
0.13
B
3.15
3.00
6.5
RBCSoline
>2.85
5.5
2.70
2’
45
2.55
.0 2.40
U
U,
5
4)
a- 2.25
3.5
>
2.10
0
,
2.5
1.95
1.80
1 .5
0.5 0.05
0.045 0.055 0.065 0.075
0.085
0.095
Flow Rote, mL/min
0.15
0.25
0.35
0.45
0.55
0.65
Flow Rote. nL/min
Ag. 3. (A) Pressure head and (B) relative viscosity as a function of
the flow rate for different fluids in a 150-m-wide x 40-1an-deep
channel.
WBCs
(mean
corpuscular
volume,
87 fL vs 230-470
fL)
(12).
For distified water, the Reynolds number ranged from
17 to 126, and the experimental
results were in good
agreement
with theoretical
predictions (solid line, Fig.
3A) obtained by solving the Navier-Stokes
equations
(4). The theory predicts linear dependence between the
pressure head (SF) and flow rate (F) according to the
equation
-“w = AF,
in which w is the viscosity
of
water and A is a constant that depends only on channel
geometry.
The increase in pressure drop above theoretical predictions
at the high flow rates is due to inertial
losses and entrance length effects, which were not induded in the theory. The shear rate, which is proportional to the mrnrimum velocity in the cross-section
divided by the half-height
of the channel,
ranged
from
3086 to 39439 s’. Because of the limitations
of the
syringe pump, the more viscous fluids were not tested at
the highest flow rate; and because of the inaccuracy of
the load cell at low forces, the distilled
water was not
tested at the lowest flow rate.
Saline (15.4 mmol/L) behaved almost identically to
distified water in the channel. This implies that the flow
of liquids in these small channels is not affected by the
presence of electrolytes.
In spite of the large surface-tocross-sectional
area ratio, electrostatic
forces did not
appear to play an important
role in our experiments.
0.105
0.115
0.125
Fig. 4. (A) Pressure head and (B) relative viscosity as a function of
flow rate for distilled water, serum, and RBC suspension in saline in
a 80-Lm-wtde x 20-1inI-deep channel.
For the shear rates in our experiments
>3000 s’, all
the biological fluids exhibited the approximately
linear
relationship
2B
= -‘O,B
+
i5A1 in which PB is the
pressure head needed to drive the biological fluid and 1B
can be interpreted as the viscosity of the biological fluid.
When the flow rate F is extrapolated
to zero, then the
pressure head (AP0) is not zero as in the case of Newtonian fluids (i.e., water). This behavior is similar to
that of a Bingham
plastic (a fluid that requires finite
yield stress before it begins to flow).
The relative viscosity [PR = P/(AF)
+ ‘w]
is depicted in Fig. 3B as a function of flow rate (F). For
Newtonianfluidssuchassaline,APOB
= 0andRisa
constant.
Thus, as anticipated,
the relative viscosity of
saline was constant over the range of flow rates tested
(Fig. 3B). For biological fluids at high flow rates, 1R
FLat/Lw= a constant, whereas at very low flow rates, FLR
P/(,AF);
i.e., ILR is inversely proportional
to F,
and as F
0, ILR
infinity. Thus for serum, plasma,
and cell suspensions,
the relative viscosity
increased
markedly
as the flow rate decreased (<0.35 mI.dmin);
e.g., viscosity of whole blood increased more than twofold as flow rate was reduced from 0.35 to 0.1 mL/min
(Fig. 3B).
80-pm-wide
x 20-pm-deep
channel. The relation between flow rate and relative viscosity and pressure for
distified water, serum, and RBCs in saline is depicted in
Fig. 4A and B. For distilled water, the Reynolds number
varied from 20 to 50 and the shear rate in the channel
-
-
-
-
CLINICALCHEMISTRY,Vol. 40, No. 1, 1994 45
ranged from 20 800 to 105 257 s’. The biological fluids
required a greater pressure than did distilled water to
achieve the same flow rate (Fig. 4A). Compared
with
that for the larger channel, the pressure required to
achieve a given flow rate was much higher; e.g., a serum
flow rate of 0.09 mL/min required a pressure of 1100
kPa in the small channel and 150 kPa in the large
channel. Both serum and RBCs in saline exhibited nonNewtonian
behavior (i.e., the relative viscosity
increased as the flow rate decreased) (Fig. 4B). Flow of
whole blood in this channel was not studied because of
clogging of the channel.
Albumin-coated
150-pm-wide
x 40-pm-deep channels. We also studied the flow of WBCs suspended
in
isotonic saline (1.8 x 106 cells/L) in a 150-pm-wide x
40-pm-deep
channel and through an identical channel
coated with albumin. The coating was uneven and we
were not able to measure its thickness. However, after
viewing the coating under a microscope, we estimated
that the coating’s peaks were no more than 5% of the
channel depth. We detected no difference in the pressure
drop vs flow rate measurements
of the WBCs flowing
through the coated and uncoated channels, nor did we
observe any tendency for the WBCs to attach to the
albumin (they tended to stay away from the walls).
Microfilters.
Filters of differing
designs were fabricated by reactive ion etching within channels
to study
separation
and (or) isolation of formed elements (e.g.,
blood cells). The ability of individual ifiter designs to
remove particulate material from test solutions was investigated with suspensions
of beads and RBCs. The
5-pin filter shown in Fig. 5 comprises a row of 12 complex posts that form a barrier across the channel. This
filter effectively separated
latex microbeads
(5.78 pm
diameter).
It also separated RBCs, but the separation
was less complete because of the deformability
of the
cells that allowed some cells to pass through
the filter.
We did not observe any destruction
of RBCs in the
channels, but in other studies (not reported here) conducted at higher pressures and flow rates than the ones
used here, we did observe significant RBC deformation.
Ag. 5. Filtration of 5.78-am-diameter beads ma 500-tm-wide channel by a filter with 5-tni spacings.
46
CUNICAL CHEMISTRY, Vol. 40, No. 1, 1994
DIscussIon
Whole blood is a shear-thinning
ity decreases
as the shear rate
(the apparent viscosnon-Newtonian fluid. Blood viscosity depends on cell content,
plasma protein concentration, temperature,
storage conditions, health of donor, time since eating or exercising,
and testing conditions.
For healthy males, the nominal
whole blood viscosity
is in the range 3.62-5.56 mPa at
37#{176}C
and a shear rate of 230 s
(12). At shear rates
>1000 s’ andintubeswithdiameters
>22 pm,bloodis
believed to behave like a Newtonian
liquid; but at lower
shear rates, it behaves like a Bingham plastic (a fluid
that requires a finite yield stress before it begins to
flow-e.g.,
toothpaste)
(13,14). At low shear rates, blood
exhibits a small but finite yield stress that can affect the
velocity profile and pressure drop in small tubes at low
flow rates. The viscosity of whole blood and suspensions
of RBCs exhibits a dependence on the size of the tube
when the diameter is <300 pm (Fahraeus-Lindqvist
effect) (15)-e.g., the relative viscosity of normal hiimsin
blood drops by a factor of two when the diameter of the
tube is reduced from 500 to 40 pm (15). We observed the
Fahraeus-Lindqvist
effect for RBC suspensions in the
two channel sizes studied (Figs. 3B and 4B). The relative
viscosity of the RBC suspension, at a constant flow rate of
-0.05 mL/min, changed
from 4.5 to 2.6 as the channel
size was reduced from 40 x 150 pm to 20 x 80 pm.
Our results suggest that simple liquids behave both
increases)
qualitatively
and quantitatively
according
to established theories (from their macroscopic
counterparts)
in
channels
with characteristic
dimensions
>20 pm. In
spite of the large surface-to-volume
ratio (-10
m21m3),
we were unable to detect any significant
influence
on
the test results by surface forces. In contrast to simple
fluids, biological fluids exhibited
shear-thinning,
nonNewtonian
behavior. The shape of the curves in Figs. 3B
and 4B are as expected for a plastic shear-thinning
fluid.
However, we observed this behavior at shear rates between 3000 and 20000 s’, whereas in other studies
this was observed only at flow rates <1000 s’ (13, 14).
The relative ease of handling of the fluids in the
microchannels
indicates that this type of system will be
appropriate for the fluidic component of an analytical
device designed to test nanoliter-picoliter
sample volumes. We are currently
exploiting
microfabrication
technology
for the construction of inexpensive, disposable diagnostic devices for medical applications. A further application
of the microfabricated
structures
is
with experimental
analogs
for the study of transport in
complex networks and interconnected
pores. Microchannels, such as capillary
networks, play an important
role
in biological processes, but detailed study of fluid flow
has been hampered
by the lack of convenient or versatile experimental
systems. Previous flow studies have
used either direct in vivo observation
(e.g., rabbit ear, hamster cheek pouch)
thin glass tubes (18), or microchannels
thin wires in plastic or silicone rubber
fabricated structures
offer a versatile
of microvessels
(16, 17), flow in
made by casting
(19). The microalternative
to
these test systems, and recently a similar
been used to study blood rheology in arrays
system has
(number =
2600) of very short (14.4 pm) triangular
cross-section
channels
(equivalent
diameter,
6 pm) (20).
References
1 Cassaday M, Diebler H, Herron R, Pelavin M, Svenjak D,
Vlastelica D. Capsule chemistry technology for high-speed chem-
ical analysis. Clin Chem 1985;31:1453-6.
2. Kricka
LI. Chemilumineacent
and bioluminescent
techniques
[Review]. Clin Chess 1991;37:1472-81.
3. Kricka L1, Wilding P, Pfahler J, Harley J, Bau H, Zemel JN.
Liquid transport in micron and submicron channels. Soc PhotoOptical Instruin Eng 1989;1167:159-68.
4. Pfabler JN. Liquid transport in micron and submicron size
channels [Dissertation].
Philadelphia,
PA: University of Pennsylvania, 1992:lO9pp.
5. AngeII JB, Terry SC, Barth PW. Silicon micromechanical
devices. Sd Am 1983;248:44-55.
6. Rapoport SD, Reed ML, Weiss LE. Fabrication and testing of a
microdynamic
rotor for blood flow measurements. J Micromech
Microeng 1991;l:60-5.
7. Kittilsland
G, Stemme G, Norden B. A sub-micron partide
filter in silicon. Sensors Actuators 1990A21-A23:904-7.
8. Wise KD, Najafi K. Microfabrication techniques for integrated
sensors and microsystems. Science 1991;254:1335-.42.
9. Mallon J. Nanotecbnology from a micromachinist’s perspective.
In: Crandall BC, Lewis J, ads. Nanotechnology. Research and
perspectives. Cambridge, MA MIT Press, 1992:215-40.
10. Wallis 0. Direct-current
polarization
during field-assisted
glass-metal sealing. JAm Ceram Soc 1970;53:563-7.
11. Fotino M, Merson EJ, Allen FH. Micromethod
for rapid
separation of lymphocytes from peripheral blood. Ann Clin Lab Sd
1971;1:131-3.
12. Lentner C, ed Geigy scientific
tables, VoL 3. Basel: CibaGeigy, 1984:68.
13. Whitxnore RL. Rheology of the circulation. New York: Pergamon, 1968:66-70.
14. Skalak R, Ozkaya N, Skalak TC. Biofluid mechanics. Ann Rev
Fluid Mech 1989;21:167-204.
15. Middleman S. Transport phenomena
in the cardiovascular
system. New York: Wiley, 1972:299pp.
16. Schmid-Schonbein GW, Skalak R, IJsami 5, Chien S. Cell
distribution in capillary networks. Microvasc Baa 1980;19:18-44.
17. Mayrovitz HN, Rubin R. Leukocyte distribution to arteriolar
branches: dependence on microvascular blood flow. Microvasc Res
1985;29:282-94.
18. La Celle PL Alterations by leukocytes of erythrocyte flow in
microchannels.
Blood Cells 1986;12:179-89.
19. Fenton BM, Carr RT, Cokelet GR. Non-uniform
red cell
distribution
in 20 to 100 micron bifurcations.
Microvaac Rea
1985;29:103-26.
20. Kikuchi Y, Sate K, Ohki H, Kaneko T. Optically accessible
microchannels
formed in a single.crystal
silicon substrate for
studies of blood rheology.
Microvasc
Res 1992;44:226-40.
CUNICAL CHEMISTRY,Vol. 40, No. 1, 1994
47