Microchip Devices for High-Efficiency Separations - Analytical

Erin A. Redman , J. Scott Mellors , Jason A. Starkey , and J. Michael Ramsey. Analytical ..... Asmira Selimovic , Jayda L. Erkal , Dana M. Spence , R...
0 downloads 0 Views 133KB Size
Anal. Chem. 2000, 72, 5814-5819

Microchip Devices for High-Efficiency Separations Christopher T. Culbertson, Stephen C. Jacobson, and J. Michael Ramsey*

Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, Tennessee 37831-6142

We have fabricated a 25-cm-long spiral-shaped separation channel on a glass microchip with a footprint of only 5 cm × 5 cm. Electrophoretic separation efficiencies for dichlorofluoroscein (DCF) on this chip exceeded 1 000 000 theoretical plates and were achieved in under 46 s at a detection point 22.2 cm from the injection cross. The number of theoretical plates increased linearly with the applied voltage, and at a separation field strength of 1170 V/cm, the rate of plate generation was ∼21 000 plates/s. The large radii of curvature of the turns minimized the analyte dispersion introduced by the channel geometry as evidenced by the fact that the effective diffusion coefficient of DCF was within a few percent of that measured on a microchip with a straight separation channel over a wide range of electric field strengths. A micellar electrokinetic chromatography separation of 19 tetramethylrhodamine-labeled amino acids was accomplished in 165 s with an average plate number of 280 000. The minimum resolution between adjacent peaks for this separation was 1.2. Microfabricated fluidic devices (microchips) have been successfully demonstrated for a variety of electrically driven separation techniques1-4 including capillary electrophoresis (CE),5-7 micellar electrokinetic chromatography (MEKC),8,9 open-channel electrochromatography,10 and gel electrophoresis.11-14 While the (1) Kopp, M. U.; Crabtree, H. J.; Manz, A. Curr. Opin. Chem. Biol. 1997, 1, 410-419. (2) Dolnik, V.; Liu, S.; Jovanovich, S. Electrophoresis 2000, 21, 41-54. (3) Effenhauser, C. S.; Bruin, G. J. M.; Paulus, A. Electrophoresis 1997, 18, 2203-2213. (4) Jacobson, S. C.; Ramsey, J. M. In High-Performance Capillary Electrophoresis; Khaledi, M. G., Ed.; John Wiley & Sons: New York, 1998; Vol. 146, pp 613633. (5) Harrison, D. J.; Manz, A.; Fan, Z.; Luedi, H.; Widmer, H. M. Anal. Chem. 1992, 64, 1926-1932. (6) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1993, 65, 26372642. (7) Jacobson, S. C.; Culbertson, C. T.; Daler, J. E.; Ramsey, J. M. Anal. Chem. 1998, 70, 3476-3480. (8) Kutter, J. P.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1997, 69, 51655171. (9) Moore, A. W., Jr.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1995, 67, 4184-4189. (10) Kutter, J. P.; Jacobson, S. C.; Matsubara, N.; Ramsey, J. M. Anal. Chem. 1998, 70, 3291-3297. (11) Simpson, P. C.; Roach, D.; Woolley, A. T.; Thorsen, T.; Johnston, R.; Sensabaugh, G. F.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 2256-2261. (12) Effenhauser, C. S.; Paulus, A.; Manz, A.; Widmer, H. M. Anal. Chem. 1994, 66, 2949-2953. (13) Woolley, A. T.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 1134811352.

5814 Analytical Chemistry, Vol. 72, No. 23, December 1, 2000

separative performance of microchips, measured per unit length, is similar to or exceeds that found with conventional CE,6 the absolute efficiency is often lower due primarily to the shorter separation lengths and the lower voltage drops used for microchipbased experiments. As the separations performed on microchips become more challenging, higher absolute separation efficiencies are needed. For CE, this is most easily accomplished by increasing the applied voltage and hence the separation channel field strength. The field strengths on microchips, however, have reached the point where Joule heating has become a concern. An alternate method to improve the absolute separation efficiency is to increase the separation channel length while retaining the high field strength in the separation channel. To maintain the compact footprint of the microchip, however, the lengthened separation channel will need to incorporate turns that add a geometrical contribution to analyte dispersion.15,16 This excess dispersion occurs as both the migration distance and field strength vary laterally across the width of the channel in the turns. These variations result in an electric field enhanced “racetrack” effect for the individual molecules in the analyte band where the molecules on the inside wall of the turn travel not only a shorter distance but also faster than those along the outside wall. The amount of excess dispersion introduced by a turn has been quantified using a simple, one-dimensional model16 as shown in eq 1, where θ is the angle of the turn in radians, ω is the width

[ ( (

2θw 1 - exp -

2

σturn )

))]

w2vcrc 2Dθ(rc + w/2)2 24

2

(1)

of the channel, νc is the velocity of the analyte, rc is the radius of curvature along the centerline of the turn, and D is the analyte diffusion coefficient. If eq 1 is factored into the microchip design process, then microchips can be fabricated that minimize or significantly decrease the dispersion introduced by a turn. For example, decreasing the width of the channel, decreasing the analyte velocity, increasing the diffusion coefficient, or increasing the radius of curvature of the turn will all decrease the amount of dispersion introduced by a turn. Decreasing the channel width is the most effective method to minimize the dispersion introduced (14) Waters, L. C.; Jacobson, S. C.; Kroutchinina, N.; Khandurina, J.; Foote, R. S.; Ramsey, J. M. Anal. Chem. 1998, 70, 158-162. (15) Jacobson, S. C.; Hergenro¨der, R.; Koutny, L. B.; Warmack, R. J.; Ramsey, J. M. Anal. Chem. 1994, 66, 1107-1113. (16) Culbertson, C. T.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1998, 70, 3781-3789. 10.1021/ac0006268 CCC: $19.00

© 2000 American Chemical Society Published on Web 10/26/2000

by turns because of the higher order dependence of the turnintroduced dispersion on the channel width. One method to minimize the channel width and maintain cross sectional area is to fabricate channels with a high aspect ratio where the long axis is normal to the plane of the microchip.17-19 We, currently, do not have LIGA or deep reactive ion etching (DRIE) capabilities for microchip fabrication, so this approach is unavailable to us. In addition, decreasing the velocity by lowering the electric field strength in the separation channel will not help to increase the separation efficiency as the efficiency itself is linearly proportional to the electric field strength. Modifying the diffusion coefficient by changing the viscosity will also have no effect on the magnitude of the exponential term in eq 1 because proportional changes in the electrophoretic mobility and, therefore, velocity will leave the exponential term constant. Thus, the route we chose for decreasing turn-generated dispersion on glass microchips was to increase the radius of curvature of the turn. While this increases the overall footprint of the microchip, a spiral-shaped separation channel minimizes this increase while providing large radii of curvature for the turns. An alternate method for keeping the microchip footprint small while generating an essentially unlimited separation length with a high electric field strength is synchronized cyclic capillary electrophoresis (SCCE).20,21 In SCCE, analytes are electrokinetically moved around a cyclic channel structure which is fabricated on a planar glass substrate. These devices, however, suffer from three drawbacks: (1) some of the analyte is lost at the separationaccessory channel junctions, (2) the migration window, in terms of analyte mobilities continuously decreases as the separation progresses, and (3) excess analyte dispersion is generated in the corners. Efficiencies, however, of 850 000 plates for fluorescein have been generated in just over 350 s; i.e., 5 cycles around the chip.21 In this paper, we report the fabrication of a microchip with a 25-cm-long separation channel in the shape of a spiral on a 5 × 5 cm substrate. The effective diffusion coefficient of dichlorofluorescein measured on this device was within experimental error or within a few percent of the value measured on a chip with a straight channel over a wide range of electric field strengths. At the highest electric field strengths examined, efficiencies of over 1 million plates were generated in less than 46 s. The high efficiency of the device is demonstrated by performing an MEKC separation of 19 amino acids in less than 3 min. The resolution between each successive peak in this separation was >1.2. EXPERIMENTAL SECTION Reagents and Labeling Procedure. The 20 naturally occurring amino acids (Sigma Chemical Co., St. Louis, MO) and 2′-7′dichlorofluoroscein (DCF; Aldrich, Milwaukee, WI) were used as received. All of the amino acids were individually derivatized with tetramethylrhodamine isothiocyanate (TRITC; Molecular Probes Inc., Eugene, OR) by dissolving the TRITC in dimethyl sulfoxide (17) He, B.; Regnier, F. J. Pharm. Biomed. Anal. 1998, 17, 925-932. (18) He, B.; Tait, N.; Regnier, F. Anal. Chem. 1998, 70, 3790-3797. (19) Ford, S. M.; McWhorter, S.; Davies, J.; Soper, S. A.; Klopf, M.; Calderon, G.; Saile, V. J. Microcolumn Sep. 1998, 10, 413-422. (20) von Heeren, F.; Verpoorte, E.; Manz, A.; Thormann, W. Anal. Chem. 1996, 68, 2044-2053. (21) Burggraf, N.; Manz, A.; Verpoorte, E.; Effenhauser, C. S.; Widmer, H. M.; de Rooij, N. F. Sens. Actuators, B 1994, 20, 103-110.

(DMSO) and adding this solution in a 1:9 (v/v) ratio to a solution of an amino acid in 150 mM sodium bicarbonate (pH 9.0).22 The amino acids in the final labeling cocktail were at a concentration of 10 mM and in a 10-fold molar excess to the TRITC. All solutions were made using distilled deionized water from a Barnstead Nanopure system (Dubuque, IA) and then filtered through 0.45µm Acrodiscs (Gelman Sciences, Ann Arbor, MI). Microchip Design. An image of a spiral microchip similar to the ones used in this study can be found in Figure 1. The microchips used for the results presented in this paper differ from this image in that they had only a single reservoir in the center of the chip. This reservoir was placed over the interior cross marked by the dashed circle. The masks for patterning the chips were designed in-house with MiniCAD (Diehl Graphsoft, Inc., Columbia, MD) and were fabricated by HTA Photomask (San Jose, CA) using a Mask and Wafer Step and Repeat System (ASET 645 15-in. Image Repeater). The spiral shaped channel on the mask was 24.9 cm long and consisted of five semicircles with decreasing radii of curvatures1.9, 1.8, 1.7, 1.6, and 0.8 cm. Each semicircle was a 36-sided polygon with 5° angles between each straight segment (Figure 1, inset). The other channels leading to the injection cross were 200 µm wide and were tapered to only 20 µm wide 500 µm from the cross intersection (Figure 1, inset). The chips were fabricated from Hoya glass mask blanks (Hoya Corp. USA, Shelton, CT) using standard photolithographic and wet chemical etching techniques as previously described.15 A closed-channel network was formed by bonding another glass substrate over the top of the etched structure. Access holes to the channels were ultrasonically drilled (Sonic-Mill, Albuquerque, NM) into one of the substrates prior to bonding. Small fluid reservoirs (140-µL capacity) were attached with Epo-tek 353ND epoxy (Epoxy Technologies, Inc., Billerica, MA) at points where the access holes were drilled. Channel widths were measured using a stylus-based surface profiler (P-10; Tencor, Mountain View, CA) and are reported as the width at half the channel depth. The spiral-shaped separation channels for the chips used in this paper were 15 µm deep, 40 µm wide in the narrow channels, and 219 µm wide in the wide channels. In addition to the spiral chip, a simple cross chip was used in some of the experiments below. This chip was also fabricated using a Hoya mask blank and had 15-µm-deep channels. Apparatus. All experiments were performed using a singlepoint detection system similar to that previously described.23 Laserinduced fluorescence (LIF) detection of the analytes was performed using either the 488- (DCF) or 514-nm (TRITC-labeled amino acids) line of a Coherent Innova 90 argon ion laser (Coherent, Inc., Palo Alto, CA) as the excitation source at a power of 8 mW. The laser beam was focused through a planoconvex lens (F ) 300 mm; Esco Products, Inc., Oak Ridge, NJ) to create an estimated excitation spot size of 50 µm on the chip. The fluorescence was collected using a 40× microscope objective (CD240-M40X; Creative Devices, Neshanic Station, NJ), and the image was focused onto a 800-µm pinhole (Melles Griot, Irvine, CA). The signal passing through the pinhole was then spectrally filtered using a 488- or 514-nm notch filter (Kaiser Optical Systems, Inc., Ann Arbor, MI) and a 530- or 580-nm band-pass filter (530df30 or (22) Haugland, R. P. Handbook of Fluorescent Probes and Research Chemicals, 6th ed.; Molecular Probes: Eugene, OR, 1996. (23) Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1996, 68, 720-723.

Analytical Chemistry, Vol. 72, No. 23, December 1, 2000

5815

Figure 1. Image of a microchip with a spiral separation channel. The reservoirs on the interior of the spiral were replaced by a single reservoir located at the injection cross (dashed circle) for the experiments reported here. The small circles are the locations of the detection pointssfrom outside to inside 0.25, 11.9, and 22.2 cm. The inset is a magnification of the injection cross and parts of the spiral channel. The buffer, sample, and sample waste channels are 219 µm wide (half-depth) for all of their length except for the last 500 µm before the injection cross where they narrow to only 40 µm (half-depth). The separation channel was 24.9 cm long and 40 µm wide (half-depth). The black arrows on the inset indicate the vertexes of the polygonally shaped channels.

580df30; Omega Optical, Brattleboro, VT) before being measured at a photomultiplier tube (PMT, 77348; Oriel Instruments, Inc., Stratford, CT). The signal from the PMT was amplified using an SR570 low noise current preamplifier (Stanford Research Systems, Inc., Sunnyvale, CA) with a 100-Hz low-pass filter. The signal from the amplifier was sampled at 100 Hz using a PCI-MIO-16XE50 multifunction I/O card (National Instruments, Inc., Austin, TX) in a G3-300 Power Macintosh. The electrophoretic separations and injections were performed using three independent and remotely programmable high-voltage (0-10 kV) power sources from a multiple source supply (2866; Bertan, Hicksville, NY). To increase the field strength in the separation channel, a negative potential was applied to the waste reservoir using a CZE2000 instrument (Spellman High Voltage Electronics Corp., Plainview, NY). The proper potentials to apply at each reservoir were determined using Kirchhoff’s rules and Ohm’s law.24 The data acquisition and high-voltage control software was written in-house using LabVIEW (National Instruments). All data analysis was performed using Igor Pro from Wavemetrics, Inc. (Lake Oswego, OR). Experimental Protocol. The DCF separations were performed in a 20 mM boric acid/100 mM TRIS buffer with the DCF at a concentration of 1-2.5 µM. The amino acid separations were performed in a 10 mM sodium borate/50 mM SDS buffer with either 20% (v/v) methanol or 10% (v/v) 2-propanol added. The labeled amino acids were diluted 1/100 in the run buffer to a concentration of 10 µM. The effective diffusion coefficient for DCF was measured dynamically (i.e., in the presence of an electric field) on chips with spiral-shaped separation channels and on chips with straight (24) Seiler, K.; Fan, Z. H.; Fluri, K.; Harrison, D. J. Anal. Chem. 1994, 66, 34853491.

5816

Analytical Chemistry, Vol. 72, No. 23, December 1, 2000

separation channels. The effective diffusion coefficient is defined as the sum of all dispersion sources that can be described by a Gaussian distribution that occur over the course of a separation. Under ideal circumstances the effective diffusion coefficient is equal to the molecular diffusion coefficient. (Giddings J. C. Unified Separation Science; John Wiley & Sons: New York 1991; p 95.) On the spiral channel chips, the measurements were made at detection points located 0.25, 11.9, and 22.2 cm from the injection cross as shown by the small circles in Figure 1. The spatial peak variance at each of these points was plotted versus its migration time. The slope of a line fitted to these points is equal to twice the diffusion coefficient. The effective diffusion coefficients in a straight channel were determined using a cross-shaped chip. The detection points on this chip were located 2.32, 2.57, 3.07, and 3.32 cm from the injection cross. In addition to the effective diffusion coefficient measurements made on the straight and spiral channels, a static diffusion coefficient measurement was also made. In this type of measurement, an analyte is injected into a straight separation channel and electromigrated a short distance into a section of the channel imaged by a CCD camera. The potentials are then removed, and images of the sample plug in the channel are taken every 5 s over a 2-min period. Axial profiles are extracted from these images, and the peak variance is calculated. The peak variance is then plotted as a function of the diffusion time, and a diffusion coefficient is determined from the slope of the line as explained above for the effective diffusion coefficient measurements. These static measurements should provide a better measure of the molecular diffusion coefficient because dispersion due to sources such as adsorption, electrodispersion, and Joule heating are eliminated or minimized.

Figure 3. Diffusion coefficient of DCF as a function of the separation channel field strength for microchips with straight and spiral channels. The square at zero field strength indicates the average value and standard deviation obtained from the static diffusion coefficient measurements.

Figure 2. (A) Number of theoretical plates and the plate height obtained for DCF as a function of the separation channel field strength on the spiral channel microchip at a detection point located 22.2 cm from the injection cross. The dashed line indicates the best fit to the theoretical plate data. (B) Electropherogram of DCF and a small contaminant. The field strength for this separation was 1170 V/cm with a detection point 22.2 cm from the injection cross. A total of 1 100 000 plates were generated for this separation.

RESULTS AND DISCUSSION Performance Evaluation. Initial experiments to test the performance of the spiral chip were conducted using a gated valve to introduce the DCF into the separation channel.25 These injections were accomplished using three positive power supplies which were connected to the sample, buffer, and sample waste reservoirs with applied potentials of 9.8, 10, and 7.5 kV, respectively. The field strength in the separation channel was varied by adjusting the voltage at the separation waste reservoir with a negative power supply from 0 to -20 kV. With this range of applied voltages, the field strength in the separation channel could be varied between 370 and 1170 V/cm. By using a negative power supply at the separation channel waste reservoir instead of grounding it, the separation channel field strength was increased by a factor of 3. The performance of the spiral chip over this range of separation channel field strengths was ascertained using DCF as the analyte (Figure 2A). The number of theoretical plates increased linearly (r ) 0.998) as expected for conditions under which diffusion is the major source of band broadening. At 1170 V/cm, an average (25) Jacobson, S. C.; Koutny, L. B.; Hergenro¨der, R.; Moore, A. W., Jr.; Ramsey, J. M. Anal. Chem. 1994, 66, 3472-3476.

of 1 040 000 ( 70 000 (n ) 6) theoretical plates was generated as DCF was separated from a small contaminant peak (Figure 2B). The average plate height for these separations was 230((10) nm, and over 20 000 plates/s were generated. Attempts to apply higher potentials to the chip, i.e., >30 kV between the buffer and separation channel waste reservoirs, resulted in arcing between the electrodes which were insulated using Tygon tubing. The electrodes were 0.999. Given the dimensions of the microchips used in these experiments, eq 1 predicts that the band broadening introduced by the turns should comprise 900 V/cm, the effective diffusion coefficients measured on the spiral chip are 16((14)% higher than those measured on the straight channel chip. However, if the 2-4% increase in the effective diffusion coefficient that is predicted by eq 1 to be introduced by the turns at these field strengths is subtracted from the effective diffusion coefficient measured for the spiral chip, then the effective diffusion coefficients for the straight and spiral channel chip are within the experimental error of each other. Although some additional dispersion is seen experimentally at the higher field strengths on the spiral channel, it is small and does not seriously degrade the added performance provided by the longer separation channel. In addition to the effective diffusion coefficient measurements made on the straight and spiral chips, a static (molecular) diffusion Analytical Chemistry, Vol. 72, No. 23, December 1, 2000

5817

Figure 4. MEKC separation of 19 TRITC-labeled amino acids in a 10 mM sodium tetraborate/50 mM SDS buffer with 20% (v/v) methanol. The field strength was 770 V/cm, and the detection point was 11.87 cm from the injection cross. The peak locations of the amino acids are indicated by their standard one-letter abbreviations.

Figure 5. MEKC separation of 19 TRITC-labeled amino acids in a 10 mM sodium tetraborate/50 mM SDS buffer with 10% (v/v) 2-propanol. The field strength was 770 V/cm, and the detection point was 11.87 cm from the injection cross. The peak locations of the amino acids are indicated by their standard one-letter abbreviations.

coefficient measurement (3.63((0.08) × 10-6 cm2/s) was also made and is plotted as a point at zero field strength on Figure 3. The effective diffusion coefficients on both the spiral and straight channel chips are higher (∼20%) than the statically measured diffusion coefficient value. While the reason for the excess dispersion found in the dynamic measurements has not been determined, analyte adsorption to the channel walls is the most probable source. In any case, this excess dispersion is not related to the turns on the spiral chip as both the straight and spiral channels are affected equally. Other sources of dispersion on the spiral chip not accounted for in the effective diffusion coefficient measurements above, such as those due to the injection plug length (160((40) µm) and detection window length (20 µm), were negligible. The plate height term calculated for the injection length averaged 10((5) nm while that for the detection window was calculated to be less than 1.2 in only 165 s. To the best of our knowledge, this is 1 order of magnitude faster than any other CE-based separation published. The migration window for the run was ∼103 s, and the peak capacity was estimated to be between 81 and 141 using the same criteria as above. The migration times for the AA peaks in the buffer containing the 2-propanol were on average 27% faster across the entire elution window, and the peak resolution and selectivity were similar to that of the buffer containing methanol. While the viscosity of the two buffers was approximately equal at 1.48 cP, the molar concentration of the organic additive was 3.5 times less in the 2-propanol-containing buffer than in the methanol-containing buffer, at 1.33 and 4.71 M, respectively. The lower organic concentration in the 2-propanolcontaining buffer presumably decreases the ζ potential less than the methanol-containing buffer and, therefore, results in a faster separation.26 The low electroosmotic mobility of the methanolcontaining buffer was also problematic for injection of the more hydrophobic compounds due to the low mobility of the micelles. For the separation performed in the 2-propanol-containing buffer, the average number of theoretical plates for D, A, Y, L, and W were 274 000, 302 000, 297 000, 275 000, and 260 000, respectively. These values were compared with those obtained using a straight channel chip with a detection length of 3.53 cm and found to be proportionally higher as expected given the differences in separation length. A plot of plate height versus field strength (Figure 6) for D, Y, and W shows that the optimum field strength for the separation in terms of efficiency is ∼970 V/cm. At this field strength, however, a significant amount of Joule heating is generated as the 10 mM sodium tetraborate/50 mM SDS with 10% (v/v) 2-propanol buffer used for the MEKC separations was ∼5.5 times more conductive than the 20 mM boric acid/100 mM TRIS buffer used for the free solution DCF separations. The conductivities of two buffers were 2370 and 438 µS, respectively. An Ohm’s law plot of the current versus the applied field strength in the separation channel is clearly nonlinear as shown in Figure 6. Although the amount of power dissipated in the separation channel was only ∼1 W/m for a field strength

5818 Analytical Chemistry, Vol. 72, No. 23, December 1, 2000

(26) Liu, Z. Electrophoresis 1999, 20, 2898-2908.

Figure 6. Plate heights of TRITC-labeled aspartic acid (D), tyrosine (Y), and tryptophan (W) as a function of the separation channel field strength. The separation channel current (9 and solid line) is also plotted on this graph as a function of the separation channel field strength. The dashed line indicates the best fit to the first three points of the current versus electric field strength data.

of 970 V/cm, the power dissipation values in the narrow parts of the sample, buffer, and sample waste channels were 8.5, 11.1, and 32.2 W/m, respectively. The nonlinearity of the i-E curve for the separation channel is probably due to the heated fluid from the buffer channel entering the separation channel and lowering the fluid viscosity. As the viscosity is lowered in this channel the ionic mobility increases, thus leading to an increase in current. Joule heating probably did not detrimentally affect the overall plate height at 970 V/cm because while the plate height term for Joule heating increased there was a decrease in the micelle sorptiondesorption plate height term. The overall plate height, however,

did begin to increase slightly for runs carried out at 1170 V/cm as the Joule heating term began to increase more rapidly. The surface temperature of the chip at this separation channel field strength was 9 °C above the ambient temperature, indicating a substantial thermal gradient inside the channel. Although the increased temperature at the higher separation field strengths, i.e., 970 and 1170 V/cm, did not adversely affect the separation efficiency, it did adversely affect the selectivity (R) of the separations. The increase in temperature decreased the distribution coefficient and, therefore, the selectivity. For example, at the highest field strength testeds1170 V/cmsall of the amino acids eluted in less than 2 min with efficiencies in excess of 500 000, but several coeluted, resulting in an overall poorer separation than at 770 V/cm. In conclusion, microchips that are designed with spiral-shaped separation channels can minimize dispersion due to turns while keeping the device footprint small. The long separation channel and high electric field strength on these chips substantially increases the separative power of microchip electrophoresis techniques as evidenced by the separation of 19 amino acids in less than 3 min. ACKNOWLEDGMENT This research is sponsored by the U.S. Department of Energy, Office of Research and Development. Oak Ridge National Laboratory is managed and operated by UT-Battelle, LLC, for the U.S. Department of Energy under Contract DE-AC05-00OR22725. The authors thank Christopher D. Thomas for fabrication of the microchips. Received for review June 1, 2000. Accepted September 20, 2000. AC0006268

Analytical Chemistry, Vol. 72, No. 23, December 1, 2000

5819