Multiple Open-Channel Electroosmotic Pumping System for

Figure 1 Diagram of the microfabricated electroosmotic pumping system. (1) open-channel electroosmotic pump, (2) micropump inlet reservoir, (3) microp...
8 downloads 0 Views 284KB Size
Download PDF
Anal. Chem. 2002, 74, 6259-6268

Multiple Open-Channel Electroosmotic Pumping System for Microfluidic Sample Handling Iulia M. Lazar* and Barry L. Karger*

Barnett Institute, Northeastern University, Boston, Massachusetts 02115

The development of a novel, fully integrated, miniaturized pumping system for generation of pressure-driven flow in microfluidic platforms is described. The micropump, based on electroosmotic pumping principles, has a multiple open-channel configuration consisting of hundreds of parallel, small-diameter microchannels. Specifically, pumps with microchannels of 1-6 µm in depth, 4-50 mm in length, and an overall area of a few square millimeters, were constructed. Flow rates of 10-400 nL/ min were generated in electric-field-free regions in a stable, reproducible and controllable manner. In addition, eluent gradients were created by simultaneously using two pumps. Pressures up to 80 psi were produced with the present pump configurations. The pump can be easily interfaced with other operational elements of a micrototal analysis system (µ-TAS) device with multiplexing capabilities. A new microfluidic valving system was also briefly evaluated in conjunction with these pumps. The micropump was utilized to deliver peptide samples for electrospray ionization-mass spectrometric (ESI-MS) detection. Microfluidics and instrument miniaturization1 have experienced significant growth in recent years. A major aspect of microfluidics refers to the manipulation of fluid flows in the microchip channels. For analytical processes, such as fluid valving,2,3 mixing,4 or electrically driven separations,2-5 flow streams are typically generated using electrical forces (electroosmotic flow or EOF); however, EOF is dependent on the surface charge of the channel walls and consequently is sensitive to the physicochemical properties of the sample (pH, ionic strength, organic content). For example, the EOF can be easily reduced or even suppressed if the channel surface is altered by contact with specific sample types. Alternatively, for some techniques, such as micro-liquid chromatography (µLC) or flow injection analysis (FIA), or for some samples, such as those containing cells that require zero electric field conditions for their manipulation, fluid flows on microchips may or must be generated by differential pressure.6-8 Typically, this is accomplished by connecting external devices to the microchip, for * Address correspondence to either author. Phone: (617) 373-2867. Fax: (617) 373-2855. E-mails: [email protected]; [email protected]. (1) Manz, A.; Graber, N.; Widmer, H. M. Sens. Actuators 1990, B1, 244-248. (2) Harrison, D. J.; Manz, A.; Fan, Z.; Lu ¨ di, H.; Widmer, H. M. Anal. Chem. 1992, 64, 1926-1932. (3) Jacobson, S. C.; Hergenro ¨der, R.; Koutny, L. B.; Warmack, R. J.; Ramsey, J. M. Anal. Chem. 1994, 66, 1107-1113. (4) Jacobson, S. C.; McKnight, T. E.; Ramsey, J. M. Anal. Chem. 1999, 71, 4455-4459. (5) Jacobson, S. C.; Hergenro ¨der, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem. 1994, 66, 2369-2373. 10.1021/ac0203950 CCC: $22.00 Published on Web 11/05/2002

© 2002 American Chemical Society

example, vacuum, gas pressure generators, or syringe pumps. However, the connection of external devices to microchips, while effective, increases the complexity of the system, and integration and multiplexing capabilities may be compromised. An approach to address the above issues is to integrate micropumping units on the microchip platform itself. A number of micropumps have already been described. Mechanical pumps that use a membrane actuated by various forces (e.g., piezoelectric, electromagnetic, pneumatic)9-15 are capable of pumping fluids of various physicochemical properties; however, the flow is pulsed, and the fabrication of the micropump is relatively complex. Nonmechanical pumps, with no moving parts, that operate on the basis of a large variety of principles (e.g., electrokinetic, centrifugal, electrohydrodynamic, etc.)16-29 have been implemented, as well. (6) McEnery, M.; Tan, A.; Alderman, J.; Patterson, J.; O’Mathuna, S. C.; Glennon, J. D. Analyst 2000, 125, 25-27. (7) Huang, Y.; Rubinsky, B. Biomed. Microdevices 1999, 2 (2), 145-150. (8) Liu, H.; Felten, C.; Xue, Q.; Zhang, B.; Jedrzjewski, P.; Karger, B. L.; Foret, F. Anal. Chem. 2000, 72, 3303-3310. (9) Franc¸ ais, O.; Dufour, I. J. Micromech. Microeng. 2000, 10, 282-286. (10) Gong, Q.; Zhou, Z.; Yang, Y.; Wang, X. Sens. Actuators, A 2000, 83, 200207. (11) Accoto, D.; Carozza, M. C.; Dario, P. J. Micromech. Microeng. 2000, 10, 277-281. (12) Laurell, T.; Wallman, L.; Nilsson, J. J. Micromech. Microeng. 1999, 9, 369376. (13) Kar, S.; McWhorter, S.; Ford, S. M.; Soper, S. A. Analyst 1998, 123, 14351441. (14) Jeong, O. C.; Yang, S. S. Sens. Actuators, A 2000, 83, 249-255. (15) Handique, K.; Burke, D. T.; Mastrangelo, C. H.; Burns, M. A. Anal. Chem. 2001, 73 (8), 1831-1838. (16) Paul, P. H.; Arnold, D. W.; Rakestraw, D. J. Proceedings of the Micro Total Analysis Systems ’98 Workshop, Banff, Canada, October 13-16, 1998; pp 4952. (17) Paul, P. H.; Arnold, D. W.; Neyer, D. W.; Smith, K. B. Proceedings of the Micro Total Analysis Systems 2000 Symposium, Enschede, Netherlands, May 14-18, 2000; pp 583-590. (18) Lazar, I. M.; Ramsey, R. S.; Jacobson, S. C.; Foote, R. S.; Ramsey, J. M. J. Chromatogr., A 2000, 892, 195-201. (19) McKnight, T. E.; Culbertson, C. T.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 2001, 73 (16), 4045-4049. (20) Zeng, S.; Chen, C.-H.; Mikkelsen, J. C., Jr.; Santiago, J. C. Sens. Actuators, B 2001, 79, 107-114. (21) Morf, W. E.; Guenat, O. T.; de Rooij, N. F. Sens. Actuators, B 2001, 72, 266-272. (22) Guenat, O. T.; Ghiglione, D.; Morf, W. E.; de Rooij, N. F. Sens. Actuators, B 2001, 72, 273-282. (23) McBride, S. E.; Moroney, R. M.; Chiang, W. Proceedings of the Micro Total Analysis Systems ’98 Workshop, Banff, Canada, October 13-16, 1998; pp 4548. (24) Jang, J.; Lee, S. S. Sens. Actuators, A 2000, 80, 84-89. (25) Lemoff, A. V.; Lee, A. P. Sens. Actuators, B 2000, 63, 178-185. (26) Sammarco, T. S.; Burns, M. A. J. Micromech. Microeng. 2000, 10, 42-55. (27) Prins, M. W. J.; Welters, W. J. J.; Weekamp, J. W. Science 2001, 291 (12), 277-280. (28) Rife, J. C.; Bell, M. I.; Horwitz, J. S.; Kabler, M. N.; Auyeung, R. C. Y.; Kim, W. J. Sens. Actuators, A 2000, 86, 135-140.

Analytical Chemistry, Vol. 74, No. 24, December 15, 2002 6259

Figure 1. Diagram of the microfabricated electroosmotic pumping system. (1) open-channel electroosmotic pump, (2) micropump inlet reservoir, (3) micropump outlet reservoir, (4) double-T sample injection element, (5) channel for sample infusion or separation, (6) sample inlet reservoir, (7) sample waste reservoir, (8) channels for sample inlet, (9) channels for sample outlet, (10) ESI emitter. The inset shows an expanded view of the micropump outlet reservoir containing the porous glass disk. Details of the device can be found in the Experimental Section.

The flow produced by these pumps is generally pulse-free and varies from a few nanoliters per minute to hundreds of microliters per minute; however, the generated pressures are low, up to only a few psi. Fabrication procedures are often complex, as well. The use of electroosmosis for pressurized pumping was advanced almost 40 years ago30,31 and demonstrated later in open32 and packed capillary columns16,17 and for open-channel microchip platforms.18-22 While the packed electroosmotic pumps were shown to be capable of generating both flow and pressure, the open-channel configurations were capable of producing only flow, but not sufficient pressure. The goal of this research was to develop a simple, miniaturized pumping unit capable of stable fluid delivery at flow rates and backpressures compatible with common analytical applications on a microchip (i.e., 0.05-1 µL min-1 and up to 100 psi), and sufficiently small to enable multiplexing of individual pumps. A multiple channel micropump based on electroosmotic pumping principles consisting of parallel shallow microchannels that deliver fluid to a microfluidic network of channels of much larger dimensions was constructed. This pump design allows the generation of both flow and pressure, with a simple open-channel configuration and offers numerous advantages. First, it can be easily integrated on microfluidic platforms and utilized for fluid propulsion on the chip. Second, its fabrication using standard photolithographic and wet chemical etching technologies ensures high manufacturing reproducibility. Third, the simplicity of the design ensures robustness, reliability, and trouble-free operation. The electroosmotic micropump was designed to fit into an integrated microfluidic analysis scheme for delivery of peptide samples for electrospray ionization-mass spectrometric (ESI-MS) analysis. In addition, the ability to control reproducibly very low flow rates, at low- and high-pressure drops, enables the utilization (29) Duffy, D. C.; Gillis, H. L.; Lin, J.; Sheppard, N. F., Jr.; Kellogg, G. J. Anal. Chem. 1999, 71 (20), 4669-4678. (30) Rice, C. L.; Whitehead, R. J. Phys. Chem. 1965, 69 (11), 4017-4024. (31) Pretorius, V.; Hopkins, B. J.; Schieke, J. D. J. Chromatogr. 1974, 99, 2330. (32) Dasgupta, P. K.; Liu, S. Anal. Chem. 1994, 66, 1792-1798.

6260

Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

of the micropump for many other µ-TAS applications, such as sample transfer to microreactors, liquid chromatography, matrix assisted laser desorption ionization (MALDI)-MS or various other detection systems, etc. The fundamental principles behind the operation of this pump, its implementation on glass microfluidic platforms, and the evaluation of its performance are presented. EXPERIMENTAL SECTION Instrumentation and Methods. Microchip devices were fabricated from glass using photolithography and wet chemical etching procedures.3 Glass substrates were prepared from 1.6mm-thick glass slides sputtered with chrome and positive photoresist (Hoya Corporation, Shelton, CT). Micropump and sample handling channels were defined on a photomask by 2.5- and 20µm-wide lines, respectively. Substrate etching was conducted to achieve channel depths of 1-6 µm for the micropump and 2022 µm for sample handling. Holes of 0.8-1 mm in diameter were drilled into the microchip substrate to access the pump and the channels. The cover plate was thermally bonded to the substrate at 550 °C, and glass reservoirs were bonded to the chip using epoxy glue (Epotek, Epoxy Technology, Billerica, MA). The scheme of the EOF pumping system is presented in Figure 1. The pump (1) consisted of a number (1-100) of shallow microchannels that could deliver fluid to the large-diameter channel network on the chip. All pumping channels had a common inlet (2) and outlet (3) reservoir. The pumping channels were exposed only to buffer solutions and did not come in contact with the sample. A single pumping unit, or most often two parallel units, were incorporated on one device. The voltage for EOF generation was applied between reservoirs 2 and 3. In some configurations, reservoir 3 was common for both pumps. A porous glass disk (5 mm in diameter, 0.8-1 mm in width, and 40-50-Å pore size), prepared by Chang Associates (Worcester, MA), was utilized to prevent EOF leakage in the direction of the exit electrode placed in the outlet reservoir 3. The disk allowed for exchange of ions but not of bulk eluent flow.17,32-35 Although all reservoirs on the (33) Wallingford, R. A.; Ewing, A. G. Anal. Chem. 1987, 59, 1762-1766.

chip were made of glass, reservoir 3 was fabricated from a PEEK external nut. The porous glass disk was secured at the bottom of reservoir 3 with a corresponding internal nut. This arrangement provided for robustness and exchangeability of the porous glass disk. Shallow microchannels 8 and 9 allowed for the injection of a sample plug in the sample infusion channel 5 through a double-T injector, 4. Pressure in the microfluidic device was created by inserting fused-silica capillaries (0.03-30-m long, 20-200-µm i.d.) into channel 5. The flow rate was measured by monitoring the displacement of the liquid meniscus in a short capillary (10 cm) with an internal diameter that fitted the outer diameter of the pressuregenerating capillary. This design prevented errors in flow rate measurement due to filling by capillary action. Alternatively, fluid flows on the chip were visualized with a Nikon epifluorescence microscope (Melville, NY). Electrospray was generated from 10-mm-long fused-silica capillary emitters of 20-µm i.d. × 90-µm o.d. (Polymicro Technologies, Phoenix, AZ) inserted into the chip. Mass spectra were acquired with a Mariner time-of-flight (TOF) instrument (Applied Biosystems, Framingham, MA). A volatile buffer (15 mM ammonium bicarbonate) was chosen to prevent plugging of the pumping microchannels and for stable electrospray generation. A peptide mixture was prepared by digesting bovine hemoglobin with trypsin at a substrate/enzyme ratio of 5:1 in a 15 mM ammonium bicarbonate in water/methanol (80:20 v/v) buffer. This protocol allowed for a quick on-chip digestion process (15-20 min) that enabled on-line ESI-MS without intermediate processing steps.36 Reagents. Samples were prepared in high-purity solvent mixtures. Methanol (HPLC grade) was purchased from Fisher Scientific (Fair Lawn, NJ); ammonium bicarbonate, from Aldrich (Milwaukee, WI); Rhodamine 610 chloride, from Exciton (Dayton, OH); and proteins, from Sigma (St. Louis, MO). Deionized water (18 MΩ-cm) was obtained from an Alpha-Q water purifier system (Millipore, Bedford, MA). FUNDAMENTALS OF THE ELECTROOSMOTIC MICROPUMP Background. Ideally, an electroosmotic micropumping system should be capable of delivering a large range of fluid flow rates independent of the backpressure imposed by the existing network of microfluidic channels on the chip. An optimum pump design would ensure that the generated EOF creates pressurized flow in the analysis system without flow leaking out of the pump itself. One approach to accomplishing this task would be to design a pump filled with densely packed particles or porous materials that enable forward EOF generation and prevent backflow leakage as a result of the hydraulic resistance of the packing.16,17 Another approach, presented in this paper, is to design a pump with a large number of narrow and shallow open microchannels that produce EOF. In this case, the multiple microchannels ensure the generation of sufficient flow rate, while the small dimensions of the (34) O’Shea, T. J.; Greenhagen, R. D.; Lunte, S. M.; Lunte, C. E.; Smyth, M. R.; Radzik, D. M.; Watanabe, N. J. Chromatogr. 1992, 593, 305-312. (35) Hu, S.; Wang, Z.-L.; Li, P.-B.; Cheng, J.-K. Anal. Chem. 1997, 69, 264-267. (36) Lazar, I. M.; Ramsey, R. S., Ramsey, J. M. Anal. Chem. 2001, 73, 17331739.

microchannels result in the necessary hydraulic resistance to pressurized back flow leakage. In contrast, a single, large-diameter open-channel electroosmotic micropump could generate flow only if the backpressure is small, and the flow would be highly dependent on the backpressure.37 Properly designed, multiplechannel pump configurations with small-diameter microchannels overcome this shortcoming, being capable of generating sufficient flow that is independent of backpressure. Electroosmotic Flow Generation in Single Channels. To illustrate the principle of the open-channel electroosmotic micropump, we will consider an ideal system composed of cylindrical capillaries and examine the conditions that must be met for proper pumping. For this discussion, the contributions of local hydraulic resistances to the total pressure drop will be neglected. The standard equations that describe fluid flow and velocity in pressure driven (F∆p and v∆p) and electroosmotic driven (Feof and veof) open capillary systems are as follows:38

F∆p )

π ∆p 4 d 128η L

(1)

v∆p )

1 ∆p 2 d 32η L

(2)

Feof )

π0rζ U 2 d 4η L

(3)

0rζ U η L

(4)

veof )

where ∆p ) the pressure drop across a capillary of diameter d and length L, η ) the viscosity of the fluid, 0 ) the electrical permittivity of the vacuum, r ) the relative permittivity of the medium (or dielectric constant), ζ ) the zeta potential at the capillary wall, and U ) the voltage applied across the capillary of length L. From eqs 1 and 3, it can be seen that F∆p (Poisseuille flow) and Feof (electroosmotic flow) are dependent on d4 and on d2, respectively, as a result of the fact that v∆p is dependent on d2, while veof is independent of d. Therefore, if fluid flow generation and distribution in a microfluidic structure occurs by both pressure and electroosmotic mechanisms, we can expect net fluid flow to be induced in preferential directions. Consider a system composed of a single narrow channel (I) with dimensions d1 and L1 connected to a large channel (II) with dimensions d2 and L2 (Figure 2A). If we apply a potential drop only to the narrow channel to generate flow through an electroosmotic mechanism (i.e., ∼d12), we can redistribute the flow in both channels through a pressure-controlled mechanism (i.e., ∼d14 and d24). Pressure generation is due to the fact that the field-free capillary II acts as a restrictor for the Feof that is produced in capillary I under the influence of the electric field. The Feof will distribute into a forward flow through the large channel (F) and a back flow through the narrow channel (Fb). Pressurized fluid flow will be proportional to (d1/d2)4, regardless of the method used to produce it. Practically, for a ratio of d1/d2 of only 1:10 (L1 being equal to L2), the flow will be distributed in a ratio Fb/F of 1:10 000; (37) Parce, J. W. U.S. Patent 6012902, 2000. (38) Knox, J. H.; Grant, I. H. Chromatographia 1987, 24, 135-143.

Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

6261

Figure 2. Flow distribution in a single capillary system with various diameters. (A) Diagram of a micropump with one pumping channel. (B) F/Feof coefficients, calculated from eq 5, for a micropump with one microchannel at various d1/d2 ratios.

Figure 3. Flow distribution in a multiple channel pumping system with various diameters. (A) Diagram of a micropump with n pumping channels. (B) F/Feof coefficients, calculated from eq 6, for a micropump with microchannels n ) 1, 4, 10, 25, 100, 200, and 10 000, at d1/d2 ) 0.1.

i.e., essentially all the flow will be directed through the large channel. When both pressure and a potential gradient are applied to a capillary, the resulting flow is a sum of the Poiseuille and electroosmotic flows.30 By balancing the electroosmotic input flow (Feof), given by eq 3, with the pressurized output flows (F and Fb), given by eq 1, it can be shown that the pressurized forward flow (F) in the large channel is dependent on the EOF in the narrow channel (Feof), according to the following equation:

L1 L2

F ) Feof L1 d1 + L2 d2

()

4

(5)

The F/Feof ratio is dependent on channel dimensions but independent of pressure or the actual EOF in the system. F/Feof could be defined as an efficiency coefficient for a given pump, since it will determine what fraction of the originally produced EOF is actually pumped forward in the system. F/Feof values can be calculated from eq 5 and represented as a function of d1/d2 and L1/L2 (Figure 2B). As seen in this figure, small-diameter and long pumping channels (i.e., small d1/d2 and large L1/L2 ratios), result in large F/Feof values and a more efficient electroosmotic pump. For example, for d1/d2 ) 0.1, we create a relatively effective pump, since the F/Feof coefficient decreases from 1 to only 0.9 even when the large channel is very long and imposes a considerable backpressure (L1/L2 ) 0.001). 6262

Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

Electroosmotic Flow Generation in Multiple Channels. For very small d1 values, the actual EOF in a channel and, consequently, the overall flow in the system will be quite low. For a channel diameter of 1 µm, the EOF will be 0.17 nL/min (calculated from eq 3 for a capillary with L1 ) 0.03 m and U ) 3000 V, and for parameters given in ref 38: o ) 8.85 × 10-12 C2 N-1 m-2, r ) 80, ζ ) 50 × 10-3 V, and η ) 0.001 N s m-2). To compensate for this low EOF, multiple small-diameter channels (n) can be connected to one large channel (Figure 3A). However, in this case, by balancing again the flows, F/Feof will be dependent on n.

L1 L2

F ) Feof L1 d1 +n L2 d2

()

4

(6)

The larger the n, the greater the total Feof, but F/Feof becomes smaller. F/Feof values are represented as a function of n and L1/ L2 for d1/d2 ) 0.1 in Figure 3B. For example, for d1/d2 ) 0.1 and L1/L2 ) 0.001, the F/Feof ratio drops from 0.9 to 0.09 when n is increased from 1 to 100. The optimum number (n) and dimensions (d1 and L1) of the micropump channels that can generate the desired flow in the main channel can be calculated. If we determine F/Feof from eq 6 and multiply this value with the total EOF value calculated according to eq 3, diagrams that show the total pressure-driven

Figure 4. Total forward flow generated in a pumping system with d1/d2 ) 0.1 as a function of micropump channel diameter. Number of pumping channels: n ) 1, 4, 10, 25, 100, 200, and 10 000. (A) L1/L2 ) 0.1; (B) L1/L2 ) 0.001.

Figure 5. Achievable pressure as a function of micropump channel diameter in a system with the following parameters: L1 ) 0.03 m, L2 ) 30 m, U ) 3000 V. (A) d2 ) 30 µm; (B) d2 ) 50 µm. Table 1. Variation of Calculated F/Feof Coefficients with d1/d2 and na d1/d2 1 0.1 0.01 a

n)1 0.001 0.909 0.999

n ) 10 10-04

1× 0.5 0.999

n ) 100

n ) 1000

10-5

10-6

1× 0.091 0.999

1× 0.009 0.990

n ) 10000 1 × 10-7 0.001 0.910

L1/L2 ) 0.001.

flow for given system characteristics (in this case, d1/d2 ) 0.1) can be constructed (Figure 4). As expected, the total flow rate increases with the increase of d1 and n. However, it is important to observe that when the backpressure increases significantly (i.e., L2 becomes much larger than L1), adding additional microchannels does not bring any further benefit for increasing the flow rate (compare Figure 4B and 4A). At high backpressures, the flow cannot be pumped, but rather will leak backward in proportion to the number of channels, n. The beneficial effect of increasing the number of microchannels thus requires properly chosen dimensions. At progressively smaller d1/d2 ratios, the backflow leakage becomes negligible. For example, as shown in Table 1, the F/Feof coefficients maintain a high and relatively constant value for a large range of pumping microchannels when d1/d2 drops to a value of 0.01. On the basis of the above considerations, and

depending on the specific applications, micropumps with 1001000 pumping channels of 1-3 µm in depth should perform as effective pumping systems on microfluidic platforms. Electroosmotic Pressure Generation. The pressure that can be created in the system is calculated by balancing the input electroosmotic flow in the n narrow channels with the output pressure flow through all channels.

Ud12 L1

32norζ ∆p )

d14 d24 n + L1 L2

(7)

A plot of this pressure drop as a function of pumping channel diameter for given system characteristics is shown in Figure 5. It can be seen that the pressure depends on the number and dimensions of the pumping microchannels, the voltage applied to the pump, the ζ potential, and the relative permittivity of the medium. The pressure, however, is not dependent on the fluid viscosity. In a system with enhanced restriction, that is, d2 ) 30 µm (Figure 5A), the pressure will obviously be higher than in a system with a lower restriction, that is, d2 ) 50 µm (Figure 5B). Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

6263

Table 2. Experimentally Determined Pressure-Driven Flow Using EOF Micropumps no.

n

d1, µm

d2, µm

L1, m

L2, m

d1/d2

L1/L2

F, nL/min

RSD

1 2 3 4 5 6 7 8 9 10 11 12 13

1 4 100 100 100 100 100 100 100 100 100 100 100

27 6.1 3.8 4.2 3.9 3.9 3.9 4.2 4.2 2.9 2.9 2.9 2.5

20 27 27 20 20 20 50 200 200 200 200 200 30

0.05 0.01 0.004 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03 0.03

0.01 0.03 0.04 0.3 0.3 3.0 14.4 0.3 30 0.03 0.3 30 30

1.35 0.23 0.14 0.21 0.20 0.20 0.08 0.02 0.02 0.014 0.014 0.014 0.083

5:1 1:3 1:10 1:10 1:10 1:100 1:480 1:10 1:1000 1:1 1:10 1:1000 1:1000

120 26 130 130 100 11 57 430 290 360 320 300 17

3 5 5 7 8 11 5 6 6 5

F/Feof, theor

F/Feof, exptl

0.99 0.99 0.99

1.00 0.90 0.84

The condition of maximum pressure,

∂P )0 ∂d1

(8)

nd14 d24 ) L1 L2

(9)

is accomplished when

The absolute maximum pressure that can be generated with microchannels of a given dimension can be inferred from conditions of zero flow out from the system (i.e., the second term in the denominator of eq 7 is set equal to 0) and is dependent only on the channel diameter, applied voltage, and channel surface properties.

∆pmax )

320ξU d12

(10)

For a pump that would be used for an LC separation, once the parameters of the separation column (d2, L2) and the optimum eluent flow rate are determined, the pressure necessary to generate the given flow can be calculated, and the appropriate pump configuration (n, d1, and L1) can be chosen from diagrams shown in Figures 4 or 5. For a packed column, d2 and L2 are the dimensions of an equivalent open tubular LC column that produces the same pressure drop. Alternatively, the d24/L2 ratio in the denominator of eq 7 could be replaced by a term that takes into consideration the porosity and the permeability of the column. The microchip implementation of fully integrated micro-LC separations performed on polymeric monolithic stationary phases that are commonly performed under 100 psi39 thus becomes feasible with such pumping systems. Theoretically (see eq 10), if pressure generation were the main purpose of these micropumps and if microchip material and connecting elements would be capable of withstanding the high pressures, hundreds of bars could be produced with submicrometer-sized channel EOF pumps. RESULTS AND DISCUSSION Micropump Evaluation. Micropumping units with a variety of dimensions were designed and tested for pressure and fluid (39) Moore, R. E.; Licklider L.; Schumann, D.; Lee, T. D. Anal. Chem. 1998, 70, 4879-4884.

6264 Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

Figure 6. Photograph of a pump and inlet filter on a microchip.

flow generation, capability of eluent gradient generation, and valving for sample introduction. Pumps with 1-100 microchannels of 1-50 mm in length were constructed. After etching, the pumping microchannels were triangular in shape with a depth of 1-6 µm and a corresponding base of 11-20 µm. An equivalent diameter for these triangular channels was calculated according to the formula d ) 4A/P (where A is the area and P is the perimeter of the triangle). The dimensions of the pumping microchannels were kept sufficiently large to avoid the doublelayer overlap effect.40 The spacing between the pumping channels was 25 µm, such that the total width for a 100-microchannel pump was only 2.5 mm. Since the pumping microchannels act as a filter, clogging of some of the channels was sometimes experienced. To eliminate this effect, a short filter (100 µm in length) of the same size and shape as individual pumping channels, was introduced at both ends of the pump. Thus, plugging of the filter instead of the pumping microchannels allowed the pump to maintain its full functionality for prolonged times. A photograph of the pump and inlet filter is presented in Figure 6. An overall evaluation of the flow rates generated by various micropumps is given in Table 2. To obtain relevant experimental conditions, the diameter of the pumping channels was progressively reduced, the number of pumping microchannels was increased, and the diameter and length of the restriction capillary was varied for each pump configuration to obtain the desired restriction on the EOF. The pumps were tested for a field strength of 1000 V/cm. The results in Table 2 follow the prediction of the (40) Stevens, T. S.; Cortes, H. J. Anal. Chem. 1983, 55, 1365-1370.

theoretical model. First, a decrease in the diameter of the pumping microchannels (d1) and the d1/d2 ratio and an increase in the number of pumping channels (n), results in a more powerful pump that can deliver increased fluid flow in progressively longer restrictions (experiments 1-9). Second, as expected, for constant d1/d2 values, the flow rate decreases with the increase of the restriction L2 (experiments 5-6 and 8-9). Third, for constant and sufficiently small d1/d2 values, the flow rate becomes relatively independent of the restriction L2, maintaining an approximately constant value (experiments 10-12). Specifically, it is worth noting the results obtained for pumps with microchannel diameters d1 centering around 3 µm and d1/ d2 ) 0.014 (experiments 10-12). The experimentally determined flow rates dropped from 360 to only 320 and 290 nL/min as the restriction increased from L1:L2 ) 1:1 to L1:L2 ) 1:10 and L1:L2 ) 1:1000, respectively. Theoretically, for such a low d1/d2 ratio, the F/Feof coefficients should maintain a constant value of ∼0.99 for a large variation of the restriction length. Experimentally, the corresponding F/Feof coefficients decreased from 1 to 0.84. Even though somewhat smaller, these numbers are close to the expected theoretical value and, thus, demonstrate the viability of such a pump design. Deviations of measured from calculated F/Feof ) 0.99 values can arise from a variety of conditions, most importantly errors in the measurement of d1. The contribution of local hydraulic resistances due to turns and variations in channel diameters was also neglected. Furthermore, the original EOF generated by most of these pumps could not be accurately determined because of the interference of pressure-driven backflow. Typically, the unrestricted EOF in a single 20-30-µm-deep channel was in the 300-500 nL/min range, and in some of the 100-channel micropumps, it reached up to 1 µL/min. F/Feof values were measured only for pumps where backpressure leakage was calculated to be minimal, that is, for very small d1/d2 values, such as those in experiments 10-12. We assumed for these experiments that the flow rate measured with a short restriction (L1:L2 ) 1:1) was the actual EOF produced in the pump. The micropump with d1 ) 2.5 µm was connected to a 30-µmi.d. × 30-m-long capillary, and the pressure generation capability was determined (experiment 13). For a measured flow rate of 17 nL/min, the pressure inside the pump reached ∼80 psi. This value corresponds to the pressure required to generate the measured flow rate in the 30-m capillary connected to the chip and is close to the theoretical value of the maximum achievable pressure (90 psi) in this system, calculated by eq 7. A micropump with 1.5 µm channels was constructed, as well, but evaluation was difficult to perform as a result of the plugging of many of its channels in the bonding process of the chip. During these tests, it was observed that the electric currents measured with the porous glass disk inserted in reservoir 3, Figure 1, were generally lower than those measured in the absence of the disk. Typically, the current dropped within the first 10-15 min after turning on the pump and then remained relatively constant for the rest of the experiment (30-120 min). At low electric fields applied to the pump, that is, at low electrical currents, or in some mixed aqueous/organic buffer systems, the insertion of the porous disk made no difference in the current magnitude. This effect is believed to be induced by the resistance or charging of the glass disk and is currently under investigation.

However, since micropump flow rates were measured only after current stabilization, the electric current variations had no effect on the final performance of the pump. Gradient Generation. A preliminary testing of the micropump capability to produce solvent gradients41 was performed. Two pumps with n ) 50 microchannels each were connected in parallel and filled with buffer solution, and the inlet reservoir of one of the pumps 2A, Figure 1, was filled with a 26 µM solution of fluorescent Rhodamine 610 dye. The solvent gradient was generated by increasing the potential on reservoir 2A relative to reservoir 2B in a ratio of 1:9 to 9:1 while maintaining the potential on reservoirs 3 constant. This resulted in a progressively increased Rhodamine flow to be delivered to the microchip channels network. A photograph taken at the T intersection of the two channels connecting the pumps is given in Figure 7A. The relative contribution of the two incoming flow streams mixed in a ratio of 1:1 to the final outgoing flow can be seen at the T mixer. The area of this photograph is ∼500 × 450 µm. It is evident that complete mixing of the two streams is not attainable on such a short channel length. For small molecules, such as solvents, with a diffusion coefficient (D) of ∼1 × 10-9 m2/s,42 the diffusion across a dimension of 70-100 µm will occur in ∼5 s. Consequently, the introduction of an efficient eluent mixer element before a preconcentrator or separation channel will be necessary if the residence time in the common channel is smaller than the time necessary for mixing.43 The generation of the gradient, that is, the increase of Rhodamine concentration in the mixed flow generated by the two pumps vs time, was monitored with ESIMS. An extracted ion chromatogram for the Rhodamine ion with a m/z of 443 is shown in Figure 7B. A smooth gradient was generally easy to produce; however, sharp changes in the composition of the eluent could not be monitored with the present chip configurations. In addition, we estimate that gradients that feature very low content of either of the two solvents are possible only with pump configurations with a large F/Feof coefficient when pressurized backflow leakage is minimal. Electroosmotic Valve. Typically, sample injection in microfabricated analysis systems is accomplished electrokinetically2,3 or by using pressure gradients,44 and then sample manipulations are performed with electrokinetic forces. If sample manipulations are to be performed using pressure-driven flows, a fluid valving system is necessary to prevent leakage of the fluid flow into the sample or sample waste reservoirs (Figure 1). For the present microfabricated analysis system, valving was accomplished using very narrow channels for the sample injection and waste lines. In a pressurized analysis system, narrow injection channels45 can act as efficient valving components. As depicted in Figure 1, the sample can be infused electroosmotically from reservoir 6 to 7 through the sample and sample waste channels, 8 and 9, and a double-T injector, 4.46 The double-T injector can be an open or (41) Kutter, J. P.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1997, 69, 51655171. (42) Jakeway, S. C.; de Mello, A. J.; Russell, E. L. Fresenius’ J. Anal. Chem. 2000, 366, 525-539. (43) He, B.; Burke, B. J.; Zhang, X.; Zhang, R.; Regnier, F. E. Anal. Chem. 2001, 73, 1942-1947. (44) Zhang, B.; Foret, F.; Karger, B. L. Anal. Chem. 2001, 73, 2675-2681. (45) Zhang, C.-X.; Manz, A. Anal. Chem. 2001, 73, 2656-2662. (46) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1993, 65, 26372642.

Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

6265

Figure 7. Gradient generation with two parallel pumps. (A) Photograph of two incoming flows that merge in a T-mixer (mixing ratio 1:1). (B) Extracted ion chromatogram for the 443+ ion of Rhodamine 610 dye.

packed channel as part of a preconcentration or µ-LC system. For sample introduction and waste channels with dimensions similar to the pumping microchannels, additional pressure-driven fluid loss through these channels would be negligible. Theoretically, if the number of injection and waste channels equals the number of pumping channels, the coefficient F/Feof would be unaffected by the injection/waste channels when d1/d2 ) 0.01 and would drop to only 50% of its original value when d1/d2 ) 0.1 (L1:L2 ) 1:1000 in both cases). Consequently, a configuration with appropriate dimensions could be efficiently used for valving. We termed this new valving approach as “electroosmotic valving.” The “electroosmotic valve” is open to electroosmotic driven flows and essentially closed to pressure-driven flows. To illustrate the feasibility of the valving concept, a plug of sample containing a bovine hemoglobin digest with over 30 peptide fragments was injected in the microchip system. The initial sample concentration was 2 µM for each of the R and β chains of hemoglobin, and a total sample volume of ∼10 µL was placed in the microchip reservoir. The sample was placed in reservoir (6) and a small plug (12 nL) was loaded on the chip through the double-T injector (4) by 6266 Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

applying 4 kV to reservoir 6 and 0 kV to reservoir 7, as shown in Figure 1. The voltage in reservoirs 2 and 3 was maintained at ∼2 kV. Next, the sample plug was injected down the channel (5) by applying 4 kV to the eluent reservoirs (2) and 2 kV to reservoirs 3, 6, and 7. Thus, infusion of the sample plug was ensured by a voltage drop of 2kV applied to the pump, while electrospray generation was enabled by the voltage applied to reservoirs 3, 6, and 7. Stable ESI was generated from the microchip for hours of operation (total ion current RSD was 5-6%). A mass spectrum acquired during the elution of the plug is shown in Figure 8A, and an extracted ion chromatogram for a peptide with a m/z of 589.32+ is shown in Figure 8B. The sequence coverage for the unseparated peptide mix was ∼80%. Total sample consumption for analysis (loaded on the chip) was ∼20 pmol/hemoglobin chain, and sample injected on the channel was ∼24 fmol. These results demonstrate the viability of the multichannel design for electroosmotic valving and pumping. The integration of a packing material in the infusion channel (5) would enable LC separations on a fully integrated microchip LC system. In addition, the incorporation of an adsorbing medium in the double-T (4) injector could prove useful in preconcentrating

Figure 8. Sample injection with the electroosmotic valve. (A) TOF mass spectrum of a bovine hemoglobin tryptic digest collected during elution of a sample plug. Conditions: protein (2 µM)/trypsin, 5:1; NH4HCO3 15.4 mM (pH 8.1)/CH3OH, 80:20 (v/v); TOFMS acquisition, 0.25 spectra/s; infusion with the EOF pump (n ) 50) at ∼200 nL/min; (B) Extracted ion chromatogram for a tryptic peptide fragment with m/z 589.32+; injected volume, 12 nL; TOFMS acquisition, 0.25 spectra/s.

diluted samples and in avoiding the electrophoretic bias effect inherently associated with the use of electrokinetic forces for sample mobilization. CONCLUSIONS A stand-alone microfabricated device with an integrated multichannel EOF pump has been constructed for producing controllable pressure driven flows within microfluidic channels. If desired, the pump can be employed to deliver fluid flow for off-chip applications, as well. Micropumps with 1-100 pumping channels that produced flow rates in the range of 10-400 nL/min and developed pressures up to 80 psi were constructed. The flow rate and eluent gradients were adjusted by varying the potential drop over the pump. The experimental results followed closely the trends predicted by theoretical considerations. A new approach for microfluidic valving, that is, “electroosmotic valving,” in which sample plugs can be injected in pressurized systems, was introduced and tested for the investigation of peptide samples by MS analysis.

The implementation of microfluidic propulsion elements is necessary for the further development of a wide range of applications using microfabricated analysis systems. The pump presented in this paper can be used to deliver fluid flows in electricfield-free regions and perform sample transfer, gradient generation, or fraction collection/deposition. The utility of electroosmotic pumping systems is especially important for the generation of eluent gradients at low flow rates, where conventional systems do not perform well. A number of applications can be envisioned for such micropumps: sample infusion/deposition for ESI or MALDI-MS analysis and sample delivery to microreactors, mixers, preconcentration elements, separation elements, and FIA. In addition, improved pump configurations with up to 1000 microchannels and sufficiently small diameters could be constructed with adequate microfabrication technologies. Such pumps would have capabilities to deliver higher flow rates and pressures and could successfully be utilized in conjunction with fully integrated µ-LC systems. The integration of such pumping systems in a Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

6267

multiplexed format would allow for multiple sample analysis, with no sample carry-over, in a sequential or parallel format.

Hatsopolous Scholar Fund (I.M.L.) and NIH grant GM15847 (B.L.K.). Barnett Institute contribution number 805.

ACKNOWLEDGMENT

Received for review June 18, 2002. Accepted September 14, 2002.

We thank Jeffrey Kessilman for help with the fabrication of the microchips. This work was supported by the John N.

6268

Analytical Chemistry, Vol. 74, No. 24, December 15, 2002

AC0203950