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,
Analysis

Joseph Pfahier,2 Haim H. Bau,3 Jay N. Zemel,4 and Larry J. Krlcka5 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,

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.
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
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.

in clinical

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

5Author for correspondence. Fax 215-662-7529. ReceivedJune 16, 1993; accepted September 10, 1993.

LOAD ca

I
B
Inlet Sump SIllcon-i A
A

1=c
YRDRNER apparatus

Fig. 2. ExperImental

for observing and recording flow In

Flow Channel-\ Outlet Sump

silicon microchannels.
VDU, video display unit; VCR, videocassette recorder.

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 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

Inlet Sump

Silicon

Outlet Sump

b. Cross-SectionaView (Sect. A-A)

c. Channel Cross-Section(Sect. B-B)

channel chip.

Fig. 1. (A) Silicon chip with two straight channels (airows) (upper 60 ian wide, lower 80 cufl wide); (B) Schematic of silicon

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)an environmentally in 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

A
1 600

500 450 400

a4)

1 400

o Distilled Woter #{149}Serum A RBC in saline

p1200

350

(4 4)

1000 800 600

- 300
250
(4

a-

200

I
I
0.09 0.15 0.22 0.28 0.34 0.40 Flow Rote, mL/min 0.46 0.53 0.59 0.65

a-

400
200 0

150
100 50

0.02 0.03 0.04 0.05 0.06 0.07

0.08 0.09 Flow Rote, mL/min

0.10

0.11

0.12

0.13

0 0.03

B
3.15 3.00 RBCSoline

6.5 5.5 2 45
U
U,

>2.85 2.70
2.55

.0 2.40
3.5

5
>

a- 2.25
2.10 1.95 1.80

4)

0
,

2.5
1 .5

0.045 0.055 0.065 0.075

0.085 0.095 Flow Rote, mL/min

0.105

0.115

0.125

0.5 0.05

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)

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.

(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.

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
-

CLINICAL CHEMISTRY, ol. 40, No. 1, 1994 45 V

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 coatings 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.

DIscussIon

(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 a shear rate of 230 s and (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)

Whole blood is a shear-thinning ity decreases as the shear rate

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

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

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-

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 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

1971;1:131-3. 12. Lentner C, ed Geigy scientific

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 micromachinists perspective.

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