Microfluidic Devices Connected to Fused-Silica Capillaries with

Jun 24, 1999 - Fused-silica capillaries have been connected to microfluidic devices for capillary electrophoresis by drilling into the edge of the dev...
0 downloads 4 Views 66KB Size
Anal. Chem. 1999, 71, 3292-3296

Technical Notes

Microfluidic Devices Connected to Fused-Silica Capillaries with Minimal Dead Volume Nicolas H. Bings,† Can Wang,† Cameron D. Skinner,† Christa L. Colyer,† Pierre Thibault,‡ and D. Jed Harrison*,†

Department of Chemistry, University of Alberta, Edmonton, Alberta, Canada, T6G 2G2, and National Research Council of Canada, Ottawa, Ontario, Canada, K1A 0R6

Fused-silica capillaries have been connected to microfluidic devices for capillary electrophoresis by drilling into the edge of the device using 200-µm tungsten carbide drills. The standard pointed drill bits create a hole with a conical-shaped bottom that leads to a geometric dead volume of 0.7 nL at the junction, and significant band broadening when used with 0.2-nL sample plugs. The plate numbers obtained on the fused-silica capillary connected to the chip were about 16-25% of the predicted numbers. The conical area was removed with a flattipped drill bit and the band broadening was substantially eliminated (on average 98% of the predicted plate numbers were observed). All measurements were made while the device was operating with an electrospray from the end of the capillary. The effective dead volume of the flatbottom connection is minimal and allows microfluidic devices to be connected to a wide variety of external detectors. Glass microfluidic devices have begun to show their vast potential in the field of analysis, but there exists a need to make external fluid connections to these planar devices in order to improve their performance and versatility.1,2 The uses for external world-to-chip connections include coupling to conventional capillary electrophoresis (CE) detectors such as photometers or mass spectrometers,3-10 interconnecting between devices and introduc†

University of Alberta. National Research Council of Canada. (1) Effenhauser, C. S.; Bruin, G. J. M.; Paulus, A. Electrophoresis 1997, 18, 2203-2213 (2) Colyer, C. L.; Mangru, S. D.; Harrison, D. J. J. Chromatogr., A 1997, 781, 271-276. (3) Li, J.; Thibault, P.; Bings, N. H.; Skinner, C. D.; Wang, C.; Colyer, C.; Harrison, D. J. Anal. Chem. 1999, 71, 3036-3045. (4) Figeys, D. Proceedings of the µTAS ‘98 Workshop, Eds. Harrison, D. J.; van den Berg, A. Kluwer Publishing, Dordrecht, Netherlands, October 1316, 1998, 457-462. (5) Foret, F.; Liu, H.; Zhang, B.; Felten, C.; Karger, B. L. Proceedings of the µTAS ‘98 Workshop, Eds. Harrison, D. J.; van den Berg, A. Kluwer Publishing, Dordrecht, Netherlands, October 13-16, 1998, 35-38. (6) Figeys, D.; Gygi, S. P.; McKinnon, G.; Aebersold, R. Anal. Chem. 1998, 70, 3728-3734. (7) Figeys, D.; Aebersold, R. Anal. Chem. 1998, 70, 3721-3724. (8) Figeys, D.; Ning, Y.; Aebersold, R. Anal. Chem. 1997, 69, 3153-3160. (9) Ramsey, R. S.; Ramsey, J. M. Anal. Chem. 1997, 69, 1174-1178. ‡

3292 Analytical Chemistry, Vol. 71, No. 15, August 1, 1999

ing samples.11 Typical injected sample volumes on a microchip are 0.1-0.5 nL compared to 2-5 nL in conventional CE; thus, any connectors intended to move sample off the chip after it has been processed on-chip must have extremely low dead volume. In large part, these devices find their greatest utility in highperformance separations,1,2 so the need for low dead volume connections following a separation is significant. This paper presents a method to fabricate low-picoliter dead volume connectors. Application of the connector to a chip-based electrophoresis system coupled to an electrospray ionization mass spectrometer (ESMS) system is presented elsewhere.3 Several methods have been reported for joining capillaries to microfluidic devices, but to date they have shortcomings related to their dead volumes. Chip-based ESMS has recently become a focus of attention, but the published reports have largely used the chip as an effusion source without separation,8-10 apparently due to the lack of a good low dead volume connector. For example, Figeys et al. constructed a butt joint between a capillary and the edge of a chip but did not report the dead volume of the joint, using the capillary rather than the chip for separation.4,7 In our hands, the alignment of the capillary with the chip proved very difficult. Electrospray coupling to a MS, straight out of the butt edge of a chip, appears to produce overly large droplets that show nonoptimum electrospray performance.3,9,10 Micromachining silicon substrates is more flexible, and it is possible to produce flat structures and holes that can form a connection with minimal dead volume.12,13 Unfortunately, silicon devices are not capable of sustaining the high electric fields that glass devices exploit for efficient separations.14 To address the need, we have developed a method of connecting silica capillaries to microfluidic devices. We show that the separations achieved on chip are impaired by dead volumes as small as 0.7 nL but that lower dead volume connectors that do not impair performance can be fabricated. (10) Xue, Q.; Foret, F.; Dunayevskiy, Y. M.; Zavracky, P. M.; McGruer, N. E.; Karger, B. L. Anal. Chem. 1997, 69, 426-430. (11) Ocvirk, G.; Tang, T.; Harrison, D. J. Analyst 1998, 123, 1429-1434. (12) Spiering, V. L.; van der Moolen, J. N.; Burger, G.-J.; van den Berg A. Proceedings of Transducers ‘97, June 16-19, 1997; pp 511-514. (13) van der Moolen, J. N.; Poppe, H.; Smit, H. C. Anal. Chem. 1997, 69, 42204225. (14) Harrison, D. J.; Glavina, P. G.; Manz, A. Sens. Actuators B 1993, 10, 107116. 10.1021/ac981419z CCC: $18.00

© 1999 American Chemical Society Published on Web 06/24/1999

Figure 1. Overall experimental setup. The laser beams are focused onto the device and the light is collected with optics that are not shown. An electrospray forms in the 5.0-mm gap at the end of the 5.3-cm-long capillary when the separation voltage is applied.

Figure 2. Schematic diagram of drilling a flat-bottomed hole into glass. The bottom of the hole widens out into a “fishtail” if the flattipped drill is forced to deepen the hole beyond the depth left by the pointed drill.

EXPERIMENTAL SECTION Device and Joint Preparation. Figure 1 illustrates the overall layout of the CE-MS chip design used in this study. The devices were prepared as previously described, but with the separation channel etched 13 µm deep and 40 µm wide.15 The channels labeled a-d are 300 µm wide but narrow to 40 µm, 2 mm before the junction, shown in the inset. Crystal Bond 509 (CB) (Aremcoproducts, Valley Cottage, NY) was dyed to aid visualization by mixing one to two drops of black ink from a Staedtler Lumocolor permanent pen in approximately 1-2 mL of melted CB. The device was placed on a hot plate (80 °C), and the CB was melted into a channel access hole at position J (Figure 1). The channel segment where the device was to be cut and drilled was filled with CB by applying vacuum on side channel e (Figure 1). The device was cut perpendicular to the separation channel with a diamond saw, and the face was sanded with 220- and 600-grit silicon carbide abrasive paper. This facilitated locating the end of the separation channel and reduced the risk of the drill catching on the surface. The 200-µm (+0 to -8 µm) pointed and supplier-modified flattipped tungsten carbide drills (Tycom, Mississauga, ON, Canada) were used with 185-µm-o.d. capillaries (Polymicro Technologies, Inc., Phoenix, AZ). Alternatively, flat-tipped drills were prepared in-house by grinding the tip of the drill flat while manually rotating a fine diamond wheel. The in-house flattened drill bits appeared to produce better quality holes than the commercially prepared bits. The device was clamped vertically to a horizontal XY translation stage (Newport, Irvine, CA), to position the filled channel directly below the drill bit. Use of a 20× jeweler’s loupe and side illumination, with the drill tip about 0.2-0.5 mm above the surface of the chip, facilitated alignment. The drill (4000 rpm) was lowered until it began removing glass. The chip face was then examined to ensure that the hole was centered on the channel; if not, the chip was resanded and a new hole started. A drop of water was used to lubricate and cool the drill. The drill was allowed to bore approximately 1-2 drill diameters before it was raised. This process was repeated until a hole of suitable depth (600-800 µm) was obtained. The face of the chip was cleaned with a paper towel to remove the glass powder produced during drilling. To produce the flat-bottomed holes, the pointed drill bit was replaced with the flat-tipped drill. A fresh drop of water was placed on the device, the drill was introduced and the bottom of the hole was flattened in one step, as illustrated in Figure 2. If the flattipped drill is forced to drill beyond the end of the hole left by

the pointed drill, the bottom of the hole “fishtails”, because the flat face of the drill bit is not capable of removing glass. Providing there were no air bubbles in the hole, glass debris was removed by inverting the device in a beaker of water and allowing the glass particles to settle out over a few hours. Alternatively, a capillary (at least 25 µm smaller o.d. than the hole) was used to flush the hole. The chip was then placed on a hot plate and the CB was melted and removed via the hole with the aid of vacuum. After cooling, the CB residues were removed with reagent grade acetone (Caledon Laboratories Ltd., Georgetown, ON, Canada). The capillary end was sanded flat and square with 600- and 1200-grit silicon carbide paper; glass particles were then flushed out with water. The device was placed on a 10 × 10 cm scrap of glass and the capillary was inserted into the hole. This assembly was heated to the CB melting point (∼80 °C). A small amount of CB was applied onto the face of the joint and allowed to wick into the hole until it nearly reached the end of the capillary. The rate of flow was controlled by adjusting the hot plate temperature. The assembly was removed and the CB was frozen quickly with forced air. Electrophoresis Procedures. The device and capillary assembly was flushed with water, 0.1 M NaOH, and then the running buffer for 30 min. Morpholine buffer (50 mM, pH 9.5) prepared in deionized distilled water was used, as it is suitable for ESMS. Partially decomposed (due to age) 0.1 µM FITC-labeled arginine or phenylalanine diluted with buffer was used as a sample. The labeled amino acid, stored at 4 °C, was prepared by mixing 7.63 mM amino acid with 1.52 mM FITC (Sigma) before allowing the mixture to stand overnight at room temperature.16 The high-voltage instrumentation and fluorescence detector have been previously described, except a 25× (0.35 NA) Leitz Fluotar objective, without a pinhole, and a 530DF30 (Omega Optical, Brattleboro, VT) emission filter were used.15 The device was operated in an electrospray mode, with an air gap of ∼5 mm between the pulled capillary tip and a copper counter electrode plate. The capillary was affixed to a post ∼1 cm from the electrospray tip. The formation of a Taylor cone was observed with a microscope. Sample was injected (from reservoir a to c, Figure 1) by applying -1.5 kV for 1 s at the sample waste c with b, a, and e grounded and the counter electrode left floating. Separation was performed by applying -7 kV to the counter electrode, grounding the buffer reservoir b, floating reservoir e, and applying pushback voltages of ∼-550 V, depending on liquid

(15) Fan, Z. H.; Harrison, D. J. Anal. Chem. 1994, 66, 177-184.

(16) Cheng, Y.-F.; Dovichi, N. J. Science 1988, 242, 562-564

Analytical Chemistry, Vol. 71, No. 15, August 1, 1999

3293

Figure 3. Top image, a 185 × 50 µm capillary in the hole cut into the device by the 200-µm drill. Note the large (0.7 nL) dead volume in the frusto-conical area left by the tapered point. The separation channel is 40 µm wide. Bottom image, shows the capillary in the hole produced by the flat-tipped 200-µm drill. The large dead volume left by the tapered point is absent.

levels, to reservoirs a and c.17,18 A small volume injector (d to a) was not used in this study, and so d was left floating. Two separate 2-mW argon ion laser beams (488 nm) were focused to 40-µm-diameter spots on the capillary and device, respectively. The two fluorescence signals were collected, during separation, with separate microscopes and PMTs, one from the chip, 4.4 cm from the injector, 1 mm before the junction (J), and the second on the capillary 4.9 cm after J, 4 mm from the tip of the 5.3-cm-long capillary. Two styles of connection were tested in the experiments, one where the capillary was cemented into a hole with a conical bottom from the pointed drill bit and another one with a flat-bottomed hole (Figure 3). RESULTS AND DISCUSSION In this work, we decided to further develop the capillary-tochip interface rather than attempt to generate an electrospray directly from the edge of the chip.9,10 Utilizing an attached capillary, with a tapered tip, provides a small and readily manipulated droplet size,19,20 and the length of the attached capillary can be changed to meet changing resolution needs. Such a hybrid device also allows for the use of commercial microelectrospray interfaces that provide independent control over the (17) Jacobson, S. C.; Hergenroder, R.; Koutny, L. B. Anal. Chem. 1994, 66, 1107-1113. (18) Seiler, K.; Fan, Z. H.; Harrison, D. J. Anal. Chem. 1994, 66, 3485-3491. (19) Matthias W.; Matthias M. Anal. Chem. 1996, 68, 1-8. (20) Kelly, J. F.; Ramaley, L.; Thibault, P. Anal. Chem. 1997, 69, 51-60.

3294 Analytical Chemistry, Vol. 71, No. 15, August 1, 1999

electrospray operating parameters. The important aspects of making the low dead volume connection are discussed first, followed by an evaluation of the band broadening in the hybrid device. Capillary Coupling. For the experiments presented here, we used a drill press with no measurable runout or vibrations (model 7010, Servo Products Co., Pasadena, CA). This press produced superior holes, with less cracking along the hole wall and bottom and much less drill bit breakage, compared to holes drilled with a more modest press. With this tool, the drills remained centered on the channel, even with the larger drills (370 µm). The choice of capillaries compatible with the 200-µm size of drill was limited to 185 µm o.d. and 50 µm i.d. Unfortunately, the channel cross-sectional area was ∼450 µm2 whereas the capillary area was 1960 µm2 or ∼4.4 times larger. The 150-µm nominal o.d. drill bits produced holes smaller than the 140-150-µm capillaries available, because of the relatively large negative tolerances on the drills. The larger drill bits (370 µm) produced holes suitable for =365-µm capillaries. However, interfacing the device to the mass spectrometer, the impetus for this research, is more convenient with the smaller diameter capillaries. One of the challenges in cutting and drilling a microfluidic device after it is bonded is the risk of plugging the channel with glass particles. The channel was filled with CB to avoid plugging during drilling. The important attributes of CB are as follows: it is hard, has a low melting point, accepts dyes readily, forms free flowing chips that reduce binding on the drill, is largely insoluble in water, and yet is readily soluble in acetone. Paraffin wax was also tested, but its softness allowed particles to be pressed into the channel during drilling. Crystal Bond was also used to glue the capillary into the device because the joint could be taken apart later with gentle heating. Unlike epoxy, CB did not leach contaminants observable with fluorescence or MS.8 Unfortunately, prolonged exposure of CB to water causes it to soften and expand so that it may plug the joint. Adjustment of the hot plate temperature controlled the flow rate of the glue into the joint and ensured that it made contact only with the outside of the capillary. Even though swelling of the CB still occurred, it did not block the joint. Evaluation of Band Broadening. The effect of dead volumes is to distort the peak shape and increase band broadening. The maximum separation efficiency within a hybrid microfluidic CE system is limited by the effects of both injection and detection volumes, nonideal behavior of the injected sample, longitudinal diffusion, and any additional dead volumes. The theoretical plate number N can be expressed as

N ) L2/σ2

(1)

where L is the length of the capillary. The variance of the peak, σ2, is given by

σ2 ) linj2/12 + ldet2/12 + σni2 + 2Dtmig + σdv2

(2)

where linj is the length of the injected sample (400 µm), ldet is the detector spot size (40 µm), D is the analyte diffusion coefficient, and tmig is its migration time. The σni2 term accounts for band broadening from nonideal behavior of the injected sample and

Figure 4. Electropherograms measured with fluorescence detection on the device and on a capillary connected with a tapered bottom hole. FITC-labeled phenylalanine (0.1 µM) was used as the sample in 50 mM morpholine. Note the change in the time axis scaling.

Joule heating. The σdv2 term represents the effect of the joint’s dead volume and is of an unknown form that depends on the geometric shape of the dead volume and the electric field gradients. The extracolumn band broadening prior to the joint, σec2, was determined from the plate numbers measured on the chip (Nch), and the corresponding migration time,

σec2 ) linj2/12 + ldet2/12 + σni2 ) L2/Nch - 2Dtmig (3) with L ) 4.4 cm. As an example, the value of σec2 calculated for peak 2 in Figure 4 was 3.6 × 10-4 cm2 compared to the value predicted from the injection plug and detection lengths alone of 1.3 × 10-4 cm2. Significant extracolumn band broadening has been previously observed.15,18 The predicted plate number at the end of the attached capillary was then calculated using eq 1 and

σ2 ) 2Dtmig + σec2

(4)

using the same value of D for each component of the sample. The variance due to the junction dead volume was then estimated using

σdv2 ) L2/Ncap - 2Dtmig - σec2

(5)

where Ncap is the plate number measured on the capillary, tmig is the corresponding migration time, and L ) 9.4 cm. Values of D ) 3.9 (arginine) and 3.4 (phenylalanine) × 10-6 cm2/s were used.21 Equation 2 is applicable provided that the diffusion term accurately accounts for dispersion in the channel and capillary and that any diffusion within the junction is included in σdv2. This is probably correct since the time that the sample plug spends in the joint area is small compared to tmig. The volumetric flow rate is in the range of 2-5 nL/s while the geometric dead volume of the joint is ∼0.7 nL. The plug shaping voltages used in this experiment produce a 0.18-nL sample plug. To evaluate the band-broadening influence of the interface, simultaneous fluorescence measurements were made at two (21) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1993, 65, 263268

Figure 5. Electropherograms measured with fluorescence detection on the device and on the capillary connected with a flat-bottom hole. FITC-labeled arginine (0.1 µM) was the sample in 50 mM morpholine. Note the change in the time axis scaling.

locations on the chip-capillary assembly. From Figure 4 it is clear that the tapered hole left by the pointed drill bit (Figure 3) introduces significant band broadening. For example, the plate numbers for peak 2 went from 40 000 to 15 500 on going from the device to the capillary, even though the separation length was nearly doubled. This should be contrasted with the data that were collected from the flat-bottomed hole (Figure 3) as shown in Figure 5. In this case, the plate numbers increased significantly from the device to the capillary, as expected for the increased total length. For example, the number of theoretical plates went from 47 000 to 112 000 and from 71 000 to 117 000 for peaks 1 and 3, respectively. When the observed plate numbers are compared to the maximum expected from eqs 1 and 4, the poor performance of the pointed hole connection is extremely apparent. This connection delivers only about 16-25% of the predicted value for all observed peaks, whereas the flat-bottomed hole produced observed efficiencies in the range of 72-108% (average ) 98%) of the calculated value. These results demonstrate that there is minimal dead volume at the flat-bottomed joint between device and capillary. The large change in cross-sectional areas between the chip channel and the capillary may also introduce band broadening, but no effect was observed within the precision of this study. In the case of the pointed hole connection, the variance attributed to the 0.7 nL dead volume is on average 4.1 times larger than the contribution from diffusion and extracolumn band broadening. For example, in the attached capillary σobsd2 ) 5.7 × 10-3, σcalc2 ) 1.1 × 10-3, and σdv2 ) 4.6 × 10-3 cm2 for peak 2 in Figure 4. The geometric dead volume, and diffusion within it, must contribute to the band broadening observed, but there may also be substantial distortions in the electric field22 that could play a role in dispersing the analyte zones within the pointed hole connection. Movement of the unrestrained final centimeter of capillary, caused by Coulombic forces associated with the electrospray field, resulted in changes in background and sensitivity by as much as a factor of 2. The run-to-run peak width and migration time measurements were not strongly affected by the capillary movement. The precision of the plate number calculations was 19% RSD while the precision of migration times was 1.5%. (22) Patankar, N. A.; Hu, H. H. Anal. Chem. 1998, 70, 1870-1881.

Analytical Chemistry, Vol. 71, No. 15, August 1, 1999

3295

CONCLUSIONS This study illustrates that it is possible to use a drilling procedure to generate extremely low dead volume couplings between a microfluidic chip and an external capillary. It is striking that even a 0.7-nL physical dead volume can have a notably adverse affect on the efficiency of the separation. The results demonstrate that low dead volume connections are critical for transferring separated solutions off the chip. Yet the requirements of minimized dead volume in the flow path can be easily met by using flat-tipped drills to create flat-bottomed holes. This relatively easy technique opens the possibility of using microfluidic devices attached to standard capillaries for a wide range of applications. Additionally, this coupling method makes it possible for chips to exploit commercially available detectors, such as UV absorption detectors. Microfluidic devices that can be easily connected to commercial electrospray nebulizers would greatly expand the potential of both CE and electrospray mass spectrometry for biotechno-

3296 Analytical Chemistry, Vol. 71, No. 15, August 1, 1999

logical applications. This was the reason that the study was performed in an electrospray format. This methodology will allow us to explore the value of such hybrid instrumentation.3 ACKNOWLEDGMENT We acknowledge the financial assistance of the National Research Council Canada (NRC)sNatural Sciences and Engineering Research Council of Canada (NSERC) partnership program with the industrial support of PE/Sciex. We are grateful to the Alberta Microelectronics Corp. for device fabrication. We thank Thompson Tang and Gregor Ocvirk for insights and useful discussions.

Received for review December 29, 1998. Accepted March 25, 1999. AC981419Z