Anal. Chem. 2003, 75, 3919-3923
A Nanoinjector for Microanalysis Valeri Gorbounov,† Petr Kuban,‡ Purnendu K. Dasgupta,*,‡ and Henryk Temkin†
Department of Electrical Engineering and Department of Chemistry and Biochemistry, Texas Tech University, Lubbock, Texas 79409
We describe a simple miniature injection device that can be used for introduction of nanoliter sample volumes in microfluidic systems. The hybrid microstructure consists of two hydraulically connected parts, a pulse micropump, and a multilevel cross-flow injector. Sample injection is accomplished by creating a transient pressure pulse in the sample line by means of the solenoid-based micropump. The sample line is aligned at right angles to the main carrier flow line. The two flow channels are located in two different parallel planes. The cross section of the two channels is defined by a self-sealing aperture in an elastomer. During the pressure pulse, the sample is introduced through this aperture directly into the main flow stream. Fast impulse-based injection causes rapid mixing of the injected sample with the main flow stream. This permits simple single-line manifold micro flow injection (MFI) systems. The deformation/relaxation of the elastomer is fast and repeatable; as such, rapid serial actuations essentially result in a larger injected sample volume without significantly affecting the peak shape. In the present form, 2-40-nL samples are easily injected by single injection, and the injected volume can be chosen by system parameters. The injection repeatability as observed by a photometric detector is better than 1.2% (n ) 100). Precise and reproducible sample introduction in the nanoliter scale is a nontrivial task, especially when these need to be incorporated into a miniaturized analytical system. High-pressure rotary injectors with internal loop volumes as low as 10 nL are commercially available,1 but aside from injection volume, there is nothing miniaturized about such injectors. Moreover, minimum interconnect distances (and volumes) and special requirements for use in an electric field2 makes it difficult to integrate such injectors on small systems. Many of the microfabricated valves described in the literature consist of a deflectable membrane and a stationary valve seat. The movement of the membrane is used for closing or opening the flow path of the valve. For instance, Yang et al. fabricated a valve based on a silicone rubber membrane3 that is actuated thermop†
Department of Electrical Engineering. Department of Chemistry and Biochemistry. (1) http://www.vici.com//cval/c4.htm. (2) Liu, S.; Dasgupta, P. K. Anal. Chim. Acta 1993, 283, 739-745. (3) Yang, X.; Grosjean, C.; Yu-Chong, T. J. Microelectromech. Syst. 1999, 8, 393-402. ‡
10.1021/ac034342+ CCC: $25.00 Published on Web 06/10/2003
© 2003 American Chemical Society
neumatically. Bo¨hm at al.4 used a bi-stable electromagnetic actuator for driving a valve of similar construction. The variety of such valves is far too numerous to list; other examples appear, for instance, in refs 5 and 6. However, few of these have actually been demonstrated as functioning injectors. Chip-scale systems that do not rely on an electric field for sample introduction typically rely on a pair of microfabricated valves that manipulate flow such that a sample plug is introduced into the flow of the carrier/ reagent, similar to macroscale flow injection manifolds.7 In a singleline manifold, mixing with the main carrier stream occurs principally by axial dispersion and diffusion; in a miniature scale, with flow characterized by very small Reynold’s numbers, such mixing is not facile. Turbulent mixing is possible8 but only at high flow velocities with a zigzag channel geometry. In a typical microfluidic system, with flow velocities typically in single digit mm/s, diffusion is the only motive mixing mode.9 When a sample plug replaces part of the carrier stream perpendicular to the flow direction, the mixing occurs only at the interface between the two liquids. When a second reagent line is merged with the carrier line, mixing again occurs only at the liquid interface parallel to the flow direction. To achieve an adequate degree of mixing, a long residence time, leading to increased dispersion, may be required. From this perspective, replacement of an axial slice of a flow stream with a sample is not a good strategy for achieving mixing. Several innovative approaches have therefore been developed to improve mixing in microfluidic channels. Mixing efficiency can be greatly enhanced by introducing ridgelike structures at the bottom of the channel that generate transverse flow inside the microchannels, as demonstrated independently by Johnson et al.10 and Stroock et al.11 Elwenspoek et al.12 proposed injection of a liquid into a small volume containing the second (4) Bo ¨hm, S.; Burger, G. J.; Korthorst, M. T.; Roseboom, F. Sens. Actuators, A 2000, 80, 77-83. (5) Harrison, D. J., van den Berg, A., Eds. Proceedings of Micro Total Analysis Systems 1998; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1998. Van den Berg, A., Olthuis, W., Bergveld, P., Eds. Proceedings of Micro Total Analysis Systems 2000; Kluwer Academic Publishers: Dordrecht, The Netherlands, 2000. Ramsey, J. M., van den Berg, A., Eds. Proceedings of Micro Total Analysis Systems 2001; Kluwer Academic Publishers: Dordrecht, The Netherlands, 2001. (6) Manz, A.; Becker, H.; Microsystem Technology in Chemistry and Life Science; Springer: New York, 1999. (7) Ruzicka, J.; Hansen, E. H. Flow Injection Analysis; Wiley: New York, 1988. (8) Branebjerg, J.; Fabius, B.; Gravesen, P. Micro Total Analysis Systems 1994; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1994; pp 141151. (9) Mengeaud, V.; Josserand, J.; Girault, H. H. Anal. Chem. 2002, 74, 42794286. (10) Johnson, T. J.; Ross, D.; Locascio, L. E. Anal. Chem. 2002, 74, 45-51. (11) Stroock, A. D.; Dertinger, S. K. W.; Ajdari, A.; Mezic, I.; Stone, H. A.; Whitesides, G. M. Science, 2002, 295, 647-651.
Analytical Chemistry, Vol. 75, No. 15, August 1, 2003 3919
Figure 1. The schematic of the micropump. Left: top view of the pump showing the micromachined channels (C), holes (H), and flap valve (V). Side view of the pump showing the layered structure: (a) glass slide with channels (C), (b) glass slide with holes (H), (c) flap valve layer, (d) intermediate glass layer, (e) membrane, (f) cover glass slide.
liquid through a large number of miniature holes to obtain efficient mixing. Nevertheless, fabrication complexity is not reduced by such measures. An impulse-based injection mode is used in microinjectors13 widely deployed in ink-jet printing technology.14 Septum-based injections in gas chromatography, as well as the early practice of liquid chromatography and flow injection analysis, is also wellknown, and respectable reproducibility could be obtained in manual operations by a good analyst. In this paper, we demonstrate impulse-based nanoinjection that creates sudden and transient turbulence and good mixing. The sample is injected through a self-sealing elastomeric aperture that constitutes the crosstalk window between two fluidic channels. EXPERIMENTAL SECTION Micropump Fabrication. The micropump is shown schematically in top view (left) and side view (right) in Figure 1. For clarity, only the central part of the pump with the membrane (∼6 × 6 mm) is shown. The pump consists of a base plate made of two soda lime glass slides (microscope slides, Fisher Scientific) with inlet and outlet channels and holes, a check valve made from a flexible polymer sheet, a membrane, and a cover glass slide. In the fabrication process, first two rectangular channels (C) (each 550 µm wide, 550 µm deep, 7 mm long) were machined on a microscope slide (a) by standard machining using silicon carbide tools. Then three holes (two inlet holes with 310-µm diameters and one outlet hole with a 510-µm diameter) (H) were drilled into a second glass slide (b). The two glass slides were aligned and bonded together using a thin coat of UV-cure epoxy to form the base plate. A 100-µm-thick polyester film (laser printer transparency sheet, (c)) was bonded on the top of the base plate, again using UV-cure epoxy adhesive. The circular area (∼1.2 mm in diameter), indicated as the shaded zone, contained no adhesive. After the adhesive cured, a 1.2-mm-long cut (V) was made in the polyester sheet that allowed this portion to behave like a flap valve. (12) Elwenspoek, T.; Lammerink, T. S. J.; Miyake, R.; Fluitman, J. H. J.; J. Micromech. Microeng. 1994, 4, 227-245. (13) Tseng, F.-G.; Kim, C.-J.; Ho, C.-M. J. Microelectromech. Syst. 2002, 11 (5), 427-436, 437-447. (14) http://www.imaging.org/resources/leinkjet/part1.cfm
3920 Analytical Chemistry, Vol. 75, No. 15, August 1, 2003
Figure 2. The schematic of the injector cross: (a) small-bore glass capillary (sample line), (b) silicone tubing, (c) large-bore support glass capillary, (d) self-sealing injection aperture, (e) small-bore glass capillary (carrier line).
The work chamber of the pump began with a layer of 200-µmthick glass (microscope slide cover slip, VWR scientific) (d)) with a 5-mm drilled hole. Atop this hole, a bubble-shaped polyester membrane (e), 5.5 mm in diameter and 15.5 mm radius of curvature, was cemented with UV-cure adhesive. This defined the top of the work chamber. The bubble-shaped polyester membrane was fabricated by thermal deformation of a 120-µm-thick polyester sheet to achieve the desired geometry. This membrane was pushed by the plunger solenoid for pump action. To retain the membrane in place and to provide a base for the solenoid, a 0.5mm-thick soda lime glass with a 5.5-mm drilled hole (wafer stock, Valley Design Corp., Santa Cruz) was cemented on top of the whole assembly. Stainless steel tubes, 300-µm i.d., 550-µm o.d. (Small Parts, Miami Lakes, FL), were inserted into the micromachined glass channels on the bottom plate a and sealed in place. A solenoid (G7344, Electronic Goldmine, Scottsdale, AZ) was placed on top of the cover glass with its plunger protruding through the hole to the plastic pump membrane. The solenoid was provided with a spring for return, and its actuation displacement was 400 µm. Injection Cross Fabrication. The injection cross was fabricated from three glass capillaries and silicone tubing (Figure 2). The sample line consisted of a small-bore glass capillary (a, 280µm i.d., 750-µm o.d., 8-mm long, Drummond Scientific, Broomall, PA). A rectangular groove (700 µm wide and 375 µm deep) was cut in the middle of tube a. A sleeve tubing (b, 0.36-mm i.d., 0.71mm o.d., 6 mm long, platinum-cured high purity silicone tube, St. Gobain Performance Plastics, Beaverton MI) was swollen in hexane and slipped over the aperture made in tube a. A 6-mmlong segment of a larger tube(c, 900-µm i.d., 1.45-mm o.d., Kimble) served as a housing for the a-b assembly. A rectangular groove (750 µm wide and 350 µm deep) was cut in the middle of tube c. The a-b assembly was then inserted into c (in which it fit tightly) until the elastomer covered groove in a and the larger rectangular groove in c were in alignment. A small self-sealing aperture, d ,was created in the elastomer sleeve by a sharp-tip stainless steel needle. The aperture was estimated to be 150 µm in diameter on the basis of the needle diameter. The third glass tube, e, was an 8-mm-long segment of the same type of capillary used for a. A rectangular groove (750 µm wide and 375 µm deep) was machined in the middle of this capillary. This was then laid atop the groove in c at right angles to provide a log-cabin style joint, which was secured and covered with UVcure epoxy. The inlet/outlet connectors were made of stainless steel tubings, 150-µm i.d., 300-µm o.d., 5 mm long (Small Parts)
solution at desired flow rates. PVC tubing, 300-µm i.d., was used for connecting the pulse micropump, injector, and the detector. For on-line absorbance measurements, a variable wavelength detector, a (CV,4 ISCO, Lincoln, NE) detector equipped with a 150-µm-bore fused-silica capillary as the optical cell was used at 534 nm. Off-line spectroscopic measurements were performed with an Agilent 8453 UV-vis spectrophotometer.
Figure 3. Experimental setup schematic: (S) sample inlet, (T) timed relay/swich, (M) micropump with solenoid, (I) injector assembly, (C) carrier, (D) capillary scale absorbance detector, (R) restrictor, (W) waste. Lower left inset shows the micropump and the cross-injector (the solenoid has been removed for clarity).
that were sanded on the outside to just fit the inner diameter of the 0.28-mm-i.d. glass capillaries and secured with UV-cure epoxy. Injector Operation. Referring to Figure 3, the sample line (S) is initially filled with the sample. This can be accomplished by gravity or any type of conventional pumping or aspiration. The solenoid (V) and the pump membrane (M) are in the inactive state, for example, no voltage is applied to the solenoid. When a voltage pulse, (square wave, 500-1000 ms, 6-V amplitude) is applied, the solenoid core moves downward and compresses the membrane bubble. Solenoid actuation requires only 10-25 ms, and as such, this causes a sharp pressure pulse, and the liquid is pushed out from the work chamber through the port marked “out” (Figure 2). The total displaced volume when the membrane is compressed is calculated as the volume of the membrane spherical cap15 (1/3πh2(3r - h)), where r is the diameter of the cap (15.5 mm) and h is the cap height (0.2 mm), to be ∼2 µL. However, the total injected volume can never be this large because of expansion of the membrane bubble under pressure. The pressure causes the flap valve in the work chamber to close and the elastomeric aperture to open, and some of the liquid is injected into the carrier stream. Liquid loss releases the pressure, and the injection aperture seals itself. The exact volume injected into the carrier line depends on the pressure differential between the pump chamber and the carrier line and is e50 nL. When the voltage is turned off, the solenoid and the pump membrane regain their original position. The resulting negative pressure causes the flap valve to open, and the sample chamber is refilled with the sample. The injection sequence can be repeated rapidly, at least up to one injection every 500 ms. Supporting Information is provided with this paper that provides a more detailed description schematic of the injector and a description of its operation. Chemicals. A 1 mM solution of HCl was used as the carrier liquid. A 10 mM stock solution of neutral red (Aldrich) in 1 mM HCl was diluted with the carrier solution to the desired concentration and used as samples for injection. Instrumentation. A peristaltic pump (Dynamax, Rainin Instruments, Oakland, CA) was used for pumping the carrier (15) Spiegel, M. R. Mathematical Handbook of Formulas and Tables; McGrawHill: New York, 1968.
RESULTS AND DISCUSISON Performance. The experiments were conducted with a gravity feed of the sample; the sample container was placed at a level ∼20 cm above the injector. The sample line and the pump work chamber were filled with the 1 mM neutral red as sample. The sample line waste outlet (W, Figure 3) in this case was closed. In other cases, as stated, the sample waste outlet W can be left open, or a specific flow restriction can be imposed. The amount of sample introduced into the carrier flow increases with increased restriction on the waste line. (However, here the sample inlet line is open except for the 20-cm hydrostatic head. If the maximum amount of sample is to be injected, the inlet line also should be closed after the pump chamber is filled.) If the carrier/reagent is pumped, the injected sample volume will also decrease with increasing carrier flow because that also increases the pressure drop between the injector and the detector. Except as stated below, the carrier solution flowed at a constant rate of 17 µL/ min. To measure the injection volume, a large number of sample injections (typically 90) were made, and the detector effluent was collected for the entire period. This allowed sufficient sample for macroscopic absorbance measurements in a microcuvette. The results were interpreted from a calibration curve for 5-26 µM neutral red in a 1 mM HCl medium. The calculated injection volume was 40.8 ( 0.5 nL (n ) 3) under these operating conditions. The injected sample volume decreased with increasing carrier flow rate, Q, in the range of 8.5-105 µL/min and the carrier outlet closed. The injection volume changed linearly (r2 ) 0.9990) from 42.8 to 24.9 nL within this flow rate range.
injected volume, nL ) 44.2 - 0.1829Q (µL/min) (1)
A plot is provided as Supporting Information. The linear dependence of the injected volume on the backpressure (and thus the particular flow rate) makes it facile to calculate the injection volume if the flow rate is changed. The sample enters the carrier stream through a small aperture as a transient pulse. The mixing with the carrier stream appears to be rapid: in preliminary experiments on spectrophotometric determination of chloride based on the Hg(SCN)2 - Fe3+ chemistry,16 a fast reaction, we observed that the amount of the analytical reaction product (as measured in terms of peak area) did not change upon increasing the length of the mixing channel from several millimeters to several centimeters, indicating that sufficient mixing was attained. From microscopic observation of a 40-nL sample plug injected into a channel of 250 × 250-µm cross section with a flow rate of 17 µL/min, we estimate the plug length to be ∼6 mm; for example, ∼10-fold dispersion of the sample plug (16) Ruzicka, J.; Stewart, J. W. B.; Zagatto, E. A. G. Anal. Chim. Acta 1976, 81, 387-396.
Analytical Chemistry, Vol. 75, No. 15, August 1, 2003
3921
Figure 4. The effect of multiple pump actuation on the injected volume. The upper trace represents the electric pulse width and number applied to the solenoid during each injection. The lower trace represents the peak measured in the detector. Sample outlet was closed. Flow rate, 17.5 µL/min.
occurred. However, by the time this goes to an external detector connected by a capillary of significant length, the dispersion will increase significantly. An additional significant feature of the injector is that because of the ability of the system to restore itself rapidly, repeated, fast valve actuations are possible. The amount of sample introduced can be increased by such rapidly repeated injections that seemingly are no different from a single larger volume injection. In Figure 4, we show, as an example, the result of repeated 500-msduration injections of the same sample, spaced 500 ms apart (1-5 injections). The upper trace in Figure 4 shows the number and duration of the pulses for each corresponding injection peak below. Not only does the peak area increase linearly with the number of pulses (r2 ) 0.9980), the peak half-width is constant within a relative standard deviation (RSD) of 3.2%. As already discussed, the restriction on the sample waste outlet can be varied from fully open to fully closed, and this can also be used to control the injected sample volume. Even when the sample waste outlet is fully open (with the 20-cm hydrostatic head on the inlet side), a restriction is still present on the waste line in the form of the short segment of the 300-µm i.d. stainless steel tubing. A small but detectable sample volume (∼< 0.5 nL) is still injected. When a restriction capillary is connected to the sample line outlet, the injected amount of sample will be influenced by the inner diameter and length of the capillary. Fused-silica capillaries of inner diameter 100, 150, and 200 µm and of different lengths (1, 2, 4, 6, and 10 cm) were used as restrictors; injection volumes varied between 2 and 36 nL (Figure 5). At 4-nL injection volume, the RSD of the peak area was 3.3% (n ) 15). This includes the detector reproducibility, as well; the peaks at this level are at the limit of quantitation (S/N ∼ 10). 3922 Analytical Chemistry, Vol. 75, No. 15, August 1, 2003
Figure 5. The effect of hydrodynamic restriction capillaries on the injection volume. Flow rate, 17.5 µL/min.
Figure 6. Volume injected plotted against log of the bleed resistance as measured by L/r4.
One may think of the functioning of the restriction as a resistance to leakage. If actuating the injector is analogous to imposing a sudden charge on a capacitor which then leaks out not only through the elastomeric aperture but also through the restriction resistance, the amount that is injected over the lifetime of the injection process will decrease exponentially from the moment of the impulse as this charge bleeds of both through the injection aperture and the restrictor. According to the HaagenPoiseuille equation, the restrictor resistance, R, is proportional to L/r4, where L is the length of the restrictor and r is its internal radius. The experimental data for the injected volume is linearly related to -log R, as shown in Figure 6 (r2 ) 0.98).
Injection repeatability was evaluated by consecutive injections of 1 mM neutral red solution into 1 mM HCl at a flow rate of 17.5 µL/min with the sample waste outlet closed. The peak absorbance readout is ∼0.008 A. The RSD for a large number of injections (n ) 100) in the system was 1.2% on the basis of peak height (see Supporting Information for a plot), the detector reproducibility at these levels can be a contributor to this. Except as detector noise was approached, the reproducibility with different restriction capillaries placed at the sample outlet was very similar. During this study, we performed several thousand injections without any noticeable deterioration of the injector/pump performance.
Figure 7. (a) Calibration MFIAgram for quadruplicate injections of 0.2-1 mM neutral red. Flow rate, 17.5 µL/min and (b) expanded view of one of the peaks; note minimal tailing and asymmetry.
To adapt such an injector/analysis system for automated sequential analysis of multiple samples, the flow system is simply modified by providing aspiration on the sample waste side (e.g., with a peristaltic pump) and providing a sample changing mechanism on the sample inlet side that could be based on an autosampler or carried out manually. The sample volume required to fill the pump chamber is only several microliters, and tens of repeated injections can be made from this sample volume. The needed replacement volume is ultimately controlled by the volume of the interconnects. To demonstrate sample replacement, we show a calibration MFIA-gram of quadruplicate injection of 0.2-1 mM neutral red in Figure 7a with continuous data acquisition and periodic sample replacement. Figure 7b shows one of the 1 mM response peaks in an expanded time scale such that the symmetrical nature of the peak can be readily observed. At 10% of peak height, the asymmetry is only 1.5, strongly suggesting that axial dispersion is not the principal mode of mixing in this system.
CONCLUSIONS A new nanoinjector for low-pressure systems that include capillary electrophoresis and microfluidic applications is demonstrated. Reproducible (∼1% RSD) injections of nanoliter volumes are possible; the injected volume can be changed by regulating the hydrostatic resistance of the sample line outlet or main streamflow. The injected volume can be increased by fast repeated actuation of the injector. Impulse-based injection appears to allow better mixing with the fluid stream in which the sample is injected, as compared to carrier replacement by a sample, and investigations on mixing phenomena in this injector type are underway. The device is robust. No deterioration has been observed after thousands of injections. The incorporation of the injector as an integral part of a microanalysis system will be reported soon. ACKNOWLEDGMENT This work was supported in part by Paul Whitfield Horn Professorship funds at Texas Tech University. SUPPORTING INFORMATION AVAILABLE Detailed description schematic of the injector and a description of its operation and experimental results. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review April 3, 2003. Accepted May 5, 2003. AC034342+
Analytical Chemistry, Vol. 75, No. 15, August 1, 2003
3923
