Anal. Chem. 1998, 70, 3291-3297
Solvent-Programmed Microchip Open-Channel Electrochromatography Jo 1 rg P. Kutter,† Stephen C. Jacobson, Norio Matsubara,‡ and J. Michael Ramsey*
Chemical and Analytical Sciences Division, Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, Tennessee 37831-6142
Open-channel electrochromatography in combination with solvent programming is demonstrated using a microchip device. Channel walls were coated with octadecylsilanes at ambient temperatures, yielding stationary phases for chromatographic separations of neutral dyes. The electroosmotic flow after coating was sufficient to ensure transport of all species and on-chip mixing of isocratic and gradient elution conditions with acetonitrile-buffer mixtures. Chips having different channel depths between 10.2 and 2.9 µm were evaluated for performance, and van Deemter plots were established. Channel depths of about 5 µm were found to be a good compromise between efficiency and ease of operation. Isocratic and gradient elution conditions were easily established and manipulated by computer-controlled application of voltages to the terminals of the microchip. Linear gradients with different slopes, start times, duration times, and start percentages of organic modifier proved to be powerful tools to tune selectivity and analysis time for the separation of a test mixture. Even very steep gradients still produced excellent efficiencies. Together with fast reconditioning times, complete runs could be finished in under 60 s. One of the prominent goals of microfabricated analytical tools is to develop miniature devices capable of accepting complex samples and elucidating chemical information. Such devices must be able to perform a variety of steps in the analytical process that are normally performed in the laboratory, such as sample preparation, separation, derivatization, detection, and identification. These integrated chemical analysis tools are referred to as “labon-a-chip” devices for obvious reasons. A number of publications have emerged in recent years dealing with different aspects of analytical tasks in a miniaturized domain (e.g., refs 1-4). In particular, the great potential of microchips to favorably combine two or more such tasks on a single device has been demonstrated † Present address: Microelectronics Center, Danish Technical University, Lyngby, Denmark. ‡ Permanent address: Himeji Institute of Technology, Hyogo 671-22, Japan. (1) Manz, A.; Harrison, D. J.; Verpoorte, E.; Fettinger, J. C.; Paulus, A.; Lu ¨ di H.; Widmer, H. M. J. Chromatogr. 1992, 593, 253. (2) Manz, A.; Harrison, D. J.; Verpoorte, E.; Widmer, H. M. In Advances in Chromatography; Brown, P. R., Grushka, E., Eds.; Marcel Dekker: New York, 1993; Vol. 33, p 1. (3) Raymond, D. E.; Manz, A.; Widmer, H. M. Anal. Chem. 1994, 66, 2858. (4) Ramsey, J. M.; Jacobson, S. C.; Knapp, M. R. Nature Med. 1995, 1, 1093.
S0003-2700(97)01367-X CCC: $15.00 Published on Web 07/03/1998
© 1998 American Chemical Society
(e.g., refs 5-9). On the other hand, taking a closer look at separation mechanisms, less diversity becomes apparent here, since the majority of implementations so far rely only on differences in electrophoretic mobilities of charged analytes (e.g., refs 10-14). There have been only a few papers which have demonstrated microchip separation techniques based on chromatographic interactions, in gas phase15-17 or in liquid phase.18-21 Clearly, there is a need for further investigations in this field, especially since this is a key element to the analysis of neutral species in the microchip format, such as priority pollutants, insecticides, and many other anthropogenic substances. There are a number of possible approaches to performing chromatographic separations in the microchip format. Basically these are the same options as are available in a capillary. Fluid control and pumping may be achieved either hydrodynamically or electrokinetically. Secondary chemical equilibria, which are a prerequisite for chromatographic separation, can be created most easily by mixing additives to the electrophoretic buffer. A widely used additive is sodium dodecyl sulfate (SDS).22 Alternatively, channel walls may be coated (dynamically or permanently) with (5) Jacobson, S. C.; Hergenro ¨der, R.; Moore, A. W., Jr.; Ramsey, J. M. Anal. Chem. 1994, 66, 4127. (6) Jacobson, S. C.; Koutny, L. B.; Hergenro¨der, R.; Moore, A. W., Jr.; Ramsey, J. M. Anal. Chem. 1994, 66, 3472. (7) Fluri, K.; Fitzpatrick, G.; Chiem, N.; Harrison, D. J. Anal. Chem. 1996, 68, 4285. (8) Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1996, 68, 720. (9) Woolley, A. T.; Hadley, D.; Landre, P.; deMello, A. J.; Mathies, R. A.; Northrup, M. A. Anal. Chem. 1996, 68, 4081. (10) Harrison, D. J.; Manz, A.; Fan, Z.; Lu ¨ di, H.; Widmer, H. M. Anal. Chem. 1992, 64, 1926. (11) Harrison, D. J.; Fluri, K.; Seiler, K.; Fan, Z.; Effenhauser, C. S.; Manz, A. Science 1993, 261, 895. (12) Seiler, K.; Harrison, D. J.; Manz, A. Anal. Chem. 1993, 65, 1481. (13) Jacobson, S. C.; Hergenro¨der, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem. 1994, 66, 1114. (14) Effenhauser, C. S.; Paulus, A.; Manz, A.; Widmer, H. M. Anal. Chem. 1994, 66, 2949. (15) Terry, S. C.; Jerman, J. H.; Angell, J. B. IEEE Trans. Electron Devices 1979, ED-26, 1880. (16) Reston, R. R.; Kolesar, E. S., Jr. J. Microelectromech. Syst. 1994, 3, 134. (17) Kolesar, E. S., Jr.; Reston, R. R. J. Microelectromech. Syst. 1994, 3, 147. (18) Jacobson, S. C.; Hergenro¨der, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem. 1994, 66, 2369. (19) Moore, A. W., Jr.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1995, 67, 4184. (20) von Heeren, F.; Verpoorte, E.; Manz, A.; Thormann, W. Anal. Chem. 1996, 68, 2044. (21) Kutter, J. P.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1997, 69, 5165. (22) Terabe, S.; Otsuka, K.; Ichikawa, K.; Tsuchiya, A.; Ando, T. Anal. Chem. 1984, 56, 111.
Analytical Chemistry, Vol. 70, No. 15, August 1, 1998 3291
typical LC stationary phases.23,24 Also, the entire channel can be packed with chromatographic material bonded to supporting silica particles.25,26 Even the in situ formation of separation coatings or beds via sol-gel processes or other techniques is feasible.27-30 Depending on the analytical problem, the type of stationary phase and mobile phase can be chosen to perform separations either in normal-phase mode or in reversed-phase mode. Finally, the elution strength of the mobile phase may be kept constant throughout the separation (isocratic elution mode) or varied over time (gradient elution mode). Recently, we have reported on the combination of micellar electrokinetic chromatography (MEKC) with solvent programming on microchips.21 Because of the way microchips are designed, constructed, and operated, fluidic control is easily achieved, and manipulation schemes which are required for solvent programming can be readily implemented. Other separation mechanisms besides MEKC will also benefit from these possibilities. In this paper, we explore the feasibility of solvent programming in conjunction with open-channel electrochromatography (OCEC), which is the chip equivalent of open-tubular (electro/liquid) chromatography (OTEC/OTLC).18,24,31-34 While the latter operates with open capillaries which are coated with the stationary phase, in OCEC it is rectangular channels which are chemically modified to cause chromatographic retention. The influence of this coating on the electroosmotic flow, the importance of the channel depths for separation performance, the possibilities to establish different solvent programming conditions, and the effect of various elution conditions on a sample mixture are described. EXPERIMENTAL SECTION Chemicals. Buffer solutions were prepared by dissolving sodium tetraborate (EM Science, Gibbstown, NY) in purified water and adjusting the pH to 8.4 with hydrochloric acid (EM Science). Stock solutions (20 mM) were diluted with water and different amounts of acetonitrile (EM Science) to yield 10 mM borate buffer solutions with 15 and 50% acetonitrile at about pH 8.4. The coumarin dyes (C440, C450, C460, C480) were purchased from Exciton, Inc. (Dayton, OH) and prepared as stock solutions in methanol (EM Science) at concentrations between 750 and 2500 µM. Before injection, dilutions (between 1:50 and 1:100) in the respective run buffers were prepared. The stock solutions of the coumarin dyes were kept in bottles wrapped with aluminum foil to minimize photochemical decay. Microchip Device. The microchips were fabricated in-house from glass substrates with photolithographic and wet chemical (23) Tsuda, T.; Hibi, K.; Nakanishi, T.; Takeuchi, T.; Ishii, D. J. Chromatogr. 1978, 158, 227. (24) Swart, R.; Kraak, J. C.; Poppe, H. Trends Anal. Chem. 1997, 16, 332. (25) Vissers, J. P. C.; Claessens, H. A.; Cramers, C. A. J. Chromatogr. A 1997, 779, 1. (26) Choudhari, G.; Horva´th, C. J. Chromatogr. A 1997, 781, 161. (27) Guo, Y.; Colon, L. A. Anal. Chem. 1995, 67, 2511. (28) Guo, Y.; Colon, L. A. J. Microcolumn Sep. 1995, 7, 485. (29) Peters, E. C.; Petro, M.; Svec, F.; Frechet, J. M. J. Anal. Chem. 1997, 69, 3646. (30) Palm, A.; Novotny, M. V. Anal. Chem. 1997, 69, 4499. (31) Knox, J. H.; Gilbert, M. T. J. Chromatogr. 1979, 186, 405. (32) Knox, J. H.; Grant, I. H. Chromatographia 1987, 24, 135. (33) Bruin, G. J. M.; Tock, P. P. H.; Kraak, J. C.; Poppe, H. J. Chromatogr. 1990, 517, 557. (34) Kraak, J. C. Pure Appl. Chem. 1997, 69, 157.
3292 Analytical Chemistry, Vol. 70, No. 15, August 1, 1998
etching techniques as described elsewhere.6 Reservoirs were fitted to the ends of the channels to contain the buffer and analyte solutions. To avoid evaporation losses, caps were put over the reservoirs. Platinum leads were inserted through the caps’ septa into the solutions. The layout of the chips used is the same which has been employed for solvent mixing in MEKC as recently reported.21 The delay volume of the mixing channel is of the order of 0.5-1 nL, depending on the channel geometry. Details on channel geometries (depths and widths) are specified in Table 1 in the Results and Discussion section. Coating Procedure. The coating procedure used was based on a method described by Kohr and Engelhardt.35 The chip was prepared by rinsing with 1 M sodium hydroxide solution (J. T. Baker, Phillipsburg, NJ), purified water, 1 M hydrochloric acid, water, and methanol for 10 min each. After this, the chip was dried in an oven at 110 °C overnight. Toluene (EM Science) was dried over 3-Å molecular sieve (EM Science) for at least 5 h. With the dried toluene, a 1% (v/v) solution of n-butylamine (EM Science) was prepared. The coating solution itself was then prepared by adding approximately 10% (w/w) of the silanizing reagent to dried toluene plus a few microliters of the n-butylamine solution as catalyst (about 10 µL/g of solution). Two silanizing agents were tested in this study: octadecyltrimethoxysilane and octadecyldiisobutyl(dimethylamino)silane (Gelest, Tullytown, PA). The freshly prepared coating solution was filled into the channels by applying suction at the reservoirs, all reservoirs were filled to the rim with the same solution, and the chip was placed in a container with a saturated toluene atmosphere. After approximately 24 h at room temperature, the excess coating solution was washed out of the channels with toluene, which then was removed by rinsing with methanol. Chip Operation. Fluid control in the microchips during injection and during isocratic or solvent-programmed runs was accomplished by applying voltages to the reservoirs in an appropriate fashion. Details about the way these voltages were calculated and a typical injection and run sequence have been described recently.21 Briefly, the voltages for the two mobile-phase reservoirs feeding into the mixing tee were set to specific values for a defined mixing ratio or varied over time for gradient runs. In both cases, these manipulations were done in such a way as to leave all other voltages unchanged and ensure a constant separation field strength. Custom-made power supply boxes with four high-voltage power supplies (UltraVolt, Ronkonkoma, NY) were individually controlled by a Macintosh PowerPC 7500 (Apple, Cupertino, CA) running LabView software (National Instruments, Austin, TX). All injections were done using the gated injection technique.6 In this method, under preinjection and run conditions, a constant stream of analyte flowing into the injection cross is cut off by pure buffer, which also flows into the main channel. For injection, the voltages are switched so that the analyte stream is now entering the main channel. After the injection time has elapsed, this flow is cut off again. Typical injection times were 0.3 s, resulting in injection volumes between ∼12 pL (for 2.9-µmdeep and 23-µm-wide channels, u ) 0.58 mm/s) and ∼230 pL (for 10.2-µm-deep and 56-µm-wide channels, u ) 1.32 mm/s), depending on the linear velocity and the channel geometry. No (35) Kohr, J.; Engelhardt, H. J. Chromatogr. A 1993, 652, 309. (36) Deleted in proof.
Table 1. Effect of Coating on the Magnitude of the Electroosmotic Flow (EOF) for Chips with Different Geometries and Different Silanizing Agentsa
chip
depth (µm)
width (µm)
C-4.7 C-8.7 C-2.9 S-10.2 S-5.2 S-4.8
4.7 8.7 2.9 10.2 5.2 4.8
33 45 23 56 50 50
EOF (×10-4 cm2/(V‚s))
∆EOF (×10-4 cm2/(V‚s))
before
after
absolute
relative (%)
surface (mm2)
volume (mm3)
surface/volume (1/mm)
reagent
2.96 3.14 3.24 2.25 2.78 2.78
2.34 2.71 2.89 1.89 2.09 2.10
0.62 0.43 0.35 0.36 0.69 0.68
20.9 13.7 10.8 16.0 24.8 24.5
3.73 5.78 2.88 10.53 7.63 8.15
0.008 0.021 0.004 0.046 0.018 0.018
486 274 781 230 425 455
trimethoxy trimethoxy trimethoxy trimethoxy N-silane trimethoxy
a Trimethoxy ) octadecyltrimethoxysilane, N-silane ) octadecyldiisobutyl(dimethylamino)silane. Chips are labeled according to their layout (C ) cross-chip, S ) side-tee) and the channel depth. Buffer: 10 mM borate with 30% (v/v) acetonitrile. EOF determination by indirect fluorometry (see text).
peak shape distortion due to overloading was observed when injecting these volumes of the analyte solutions at micromolar concentrations. Detection. The coumarin dyes were detected using laserinduced fluorescence (LIF). The excitation source was a krypton ion laser (Innova 300, Coherent, Palo Alto, CA) operated in multiline mode (350.7-356.4 nm) with ∼30 mW at the chip. Plasma lines were removed with a dichroic mirror (Omega Optical, Brattleboro, VT). The laser beam was then focused onto a single point on the chip near the end of the separation channel.8 Fluorescence was collected with a 10× microscope objective (N.A. 0.30, Nikon 214442), filtered spatially with a pinhole (600 µm) and spectrally with a band-pass filter (450DF40, Omega Optical) and registered by a photomultiplier tube (model 77348, Oriel, Stratford, CT). The signal was amplified (428-MAN, Keithley, Cleveland, OH) and read into the Macintosh computer via analog-digital converter cards (PCI-MIO-16XE-50, National Instruments). RESULTS AND DISCUSSION Effect of Coating on the Electroosmotic Flow. To establish the effect of the C18 coating on the magnitude of the electroosmotic flow (EOF), measurements of this flow before and after the coating procedure were performed by indirect fluorometry as described elsewhere.18 Briefly, in this method, the electroosmotic flow velocity is determined by injecting a plug of water or buffer and measuring its arrival time at the detector. Since the run buffer is doped with fluorescent dye, the injected plug will produce a negative peak. The concentration of the fluorescent dye has to be kept as low as possible (here, about 5 µM), however, so that it does not influence the ζ-potential at the wall and, hence, the electroosmotic flow. In the coated channels, the elution time of the coumarin dye C440 coincided with the void peak obtained by indirect fluorometry, indicating very little chromatographic retention for C440. For practical reasons, the elution time of C440 was used as the EOF marker in all subsequent experiments. Table 1 lists the measured electroosmotic flows before and after coating for six chips varying in design and channel geometry. In all cases, the flow velocity was diminished to some extent (1025%) after the coating procedure. There is no apparent correlation between geometrical properties of the chips and the corresponding EOF reduction. Please note, however, that most chips were coated in different batches, and although experimental parameters remained approximately the same, slight changes in the pretreatment of the chips, in the concentration of the silane, and in the
water content of the toluene can lead to different coating yields. There is also a small possibility that the trimethoxysilane may produce polymeric phases, depending on the amount of water present during the coating procedure.37 On the other hand, nonreacted methoxy moieties will be hydrolyzed immediately under standard operating conditions (aqueous buffers), thus forming new Si-OH entities. Both of the above-described reactions are not possible when using the diisobutyl-N-silane. Nonetheless, comparing the EOF measurements for chips S-5.2 and S-4.8 (see Table 1), it appears that both coatings lead to very similar results. Bruin et al. reported almost unchanged EOF values between noncoated capillaries and capillaries treated with different coating reagents.33 Guo and Colon found moderate reductions in EOF after coating of capillaries.28 These results suggest that there are very few newly formed Si-OH groups and that they do not contribute much to the EOF; most of the EOF is still induced by wall Si-OH groups. In general, coating efficiencies between ∼20% for bulkier silanizing agents and ∼40% for less bulky agents are considered standard in coating procedures for silica particles in HPLC.38,39 Judging from the observed decrease in EOF in the channels, the coating densities are of comparable magnitude on the chips. Whether the trimethoxysilane-treated channels are covered with a monomeric or (partially) polymeric layer and how much percentage of carbon was present in the stationary phase layers were not determined experimentally. This notwithstanding, all chips coated according to the described procedure provided chromatographic retention for the investigated analytes while still maintaining sufficient electroosmotic flow to ensure transport through the channels. Although no systematic long-term stability measurements were performed, the C18-coated chips were operated at pH ∼8 for several weeks without apparent deterioration of performance. Isocratic Runs. In Figure 1, OCEC separations of a fourcomponent coumarin dye mixture under different isocratic conditions are depicted (15%, 29%, and 50% acetonitrile in the run buffer), obtained on a chip with 5.2-µm-deep channels. Increasing the amount of acetonitrile from 15% to 50% slightly reduces the electroosmotic flow, as measured by the retention time of the first(37) Sander, L. C.; Wise, S. A. In Retention and Selectivity Studies in HPLC; Smith, R. M., Ed.; Journal of Chromatography Library 57; Elsevier: Amsterdam, 1994; pp 337-369. (38) Unger K. K. Porous Silica; Journal of Chromatography Library 16; Elsevier: Amsterdam, 1979. (39) Nawrocki, J. J. Chromatogr. A 1997, 779, 29.
Analytical Chemistry, Vol. 70, No. 15, August 1, 1998
3293
diameter or the channel depth, especially for the small diffusion coefficients in liquid phase as opposed to those in gas phase. This is clearly stated by the Golay equation, which describes the plate height H for open-tubular systems:41
2Dm dc2 df2 + fm(k) u + fs(k) u H) u Dm Ds
Figure 1. Isocratic elution in open-channel electrochromatography on microchips. Channel depth, 5.2 µm; stationary phase, octadecylsilane; mobile phase, 10 mM borate buffer (pH 8.4) with different amounts of acetonitrile, (a) 15%, (b) 29%, and (c) 50%; analytes, (1) C440, (2) C450, (3) C460 and (4) C480; field strength, 500 V/cm.
eluting compound (C440). This behavior has already been observed several times in various electrophoresis experiments and is mostly attributed to changes in the viscosity and the dielectric constant of the buffer system.40 The increased acetonitrile percentage also reduces the time analytes spend in the stationary phase by shifting the partitioning equilibrium to the mobile phase. As a result of both effects, peaks will elute within a decreasing time window as the solvent strength is increased until they finally coelute at 50% acetonitrile. All three solvent conditions (15%, 29%, and 50%) were generated on-chip by mixing appropriate proportions of the solutions in the two solvent reservoirs. Moreover, all other isocratic conditions between 15% and 50% may be set in the same way. The ability to rapidly adjust solvent strength for isocratic separations is an attractive feature of these devices for solving separation problems. However, increasing the elution strength is not sufficient to deal with more complex separations. For these cases, gradient elution can be very helpful, which will be discussed below. Although elution order was found to match that determined with MEKC separations of the same coumarin mix,21 there are some aspects which clearly distinguish both separation techniques, mainly with respect to efficiency. Probably the most crucial parameter governing performance in openchannel chromatography is the channel depth. Influence of Channel Depth on Plate Height. Mass transfer in open capillaries or channels greatly depends on the capillary (40) Schwer, C.; Kenndler, E. Anal. Chem. 1991, 63, 1801.
3294 Analytical Chemistry, Vol. 70, No. 15, August 1, 1998
where Dm and Ds are the diffusion coefficients in the mobile and the stationary phases, respectively; dc is the capillary diameter or channel depth; df is the thickness of the stationary phase; u is the linear velocity of the mobile phase; and fm(k) and fs(k) are functions of the retention factor k of a solute in the mobile and the stationary phases, respectively. The function fm(k) differs slightly between capillaries and rectangular channels42 and depends more significantly on the way the mobile phase is driven through the system, i.e., by pressure or by electroosmotic flow (EOF). Because of the plug profile of the EOF, the flow itself does not contribute to the broadening of the bands, while the pressure-induced parabolic flow does. Hence, fm(k) is smaller in the electro-driven mode, resulting in smaller plate heights in open capillaries.33,34,43 van Deemter-type plots were established for a moderately retained compound (C460) on three different chips with channel depths of 8.7, 4.7, and 2.9 µm, respectively. All data points were taken at 30% acetonitrile in 10 mM borate buffer at pH ∼8.4. Since the linear velocity u in electro-driven microchip chromatography is regulated by the electric field strength E, the linear correlation between these two parameters needed to be proven first. This is demonstrated in Figure 2, where u was measured via the retention time of a nonretained compound, C440. Although there are different slopes for the three curves, indicating slight differences in the ζ-potentials, the linearity is satisfactory over the entire range, with correlation coefficients greater than 0.998. Values larger than 700 V/cm could not be applied in the experimental setup used for these experiments. Figure 3 shows the experimental determination of plate height versus mobile-phase velocity and fit with van Deemter’s equation. The retention factors of C460 were as follows on the three chips: C-8.7, k ) 0.29 ( 0.01; C-4.7, k ) 0.50 ( 0.02; C-2.9, k ) 0.55 ( 0.03. These values represent typical k values in open-tubular chromatography27,33,44 and are a reflection of the low phase ratio generally encountered in open-tubular techniques. Two main conclusions can be drawn from the results depicted in Figure 2: (a) shallower channels give smaller plate heights, as is expected from theory; (b) the minima of the curves shift toward higher linear velocities with decreasing channel depths, which means that analyses can be run faster while maintaining efficiencies. One can also observe that there is little gain in plate height when going from 4.7 to 2.9 µm for smaller linear velocities. With the more retained (and therefore broader) peaks and shallower channels, the signal-to-noise ratio becomes increasingly smaller, hindering accurate plate height measurements for those peaks. The influence of other band broadening effects, such as the injection plug length, slight sample overloading, thermal effects, or electrodis(41) Golay, M. J. E. Gas Chromatography 1958; Butterworths: London, 1958. (42) Giddings, J. C. J. Chromatogr. 1961, 5, 46. (43) Jakubetz, H.; Czesla, H.; Schurig, V. J. Microcolumn Sep. 1997, 9, 421. (44) Tan, Z. J.; Remcho, V. T. Anal. Chem. 1997, 69, 581.
Figure 2. Linear velocity, u, versus field strength, E, in open-channel electrochromatography on microchips with different channel depths: (b), 8.7 µm; (9), 4.7 µm; (2), 2.9 µm. Stationary phase, octadecylsilane; mobile phase, 10 mM borate buffer (pH 8.4) with 30% acetonitrile; analyte, coumarin C440.
Figure 3. Plate height, H, versus linear velocity, u, in open-channel electrochromatography on microchips with different channel depths: (b), 8.7 µm; (9), 4.7 µm; (2), 2.9 µm. Stationary phase, octadecylsilane; mobile phase, 10 mM borate buffer (pH 8.4) with 30% acetonitrile; analyte, coumarin C460.
persion, should also be taken into consideration. Another possible interpretation is that decreasing the channel depth from 5 to 3 µm does not significantly reduce the band broadening due to the combined effect of axial diffusion plus the sum of the abovementioned extracolumn causes. Mass transfer, which is directly proportional to u and dominant at higher values of u, can still be improved, resulting in still smaller plate heights and a minimum
at higher linear velocities. On the other hand, manufacturing and operating chips with channel depths e3 µm becomes increasingly challenging. Therefore, the experiments described below were performed using a chip with 5.2-µm-deep channels. Minimum plate heights in such a chip are about 2 µm (for an analyte with k ) 0.5), corresponding to a reduced plate height of 0.4, which compares very well with values obtained in OTEC.33,44 Gradient Runs. The application of solvent gradients in liquid chromatography can have many advantages over the isocratic mode when it comes to tuning selectivity or controlling resolution and analysis time.45 However, in capillary systems, the realization of solvent gradients has previously been technically challenging,46,47 especially with the additional requirement of avoiding the band broadening influence of pressure-driven mixing pumps.48,49 Recently, we demonstrated the versatility of the microchip approach to electrokinetically generating gradients of different forms and slopes for MEKC separations.21 In this previous work, we have shown that there is still sufficient control over the electrokinetic flow in individual parts of the microchip to adjust the amount of organic modifier in the mobile phase over time in a reproducible way. In Figure 4, the same mixture as in Figure 1 is separated using three different gradient conditions. The gradients are all linear but differ in slope, starting time, and starting percentage. It can be clearly seen how these three parameters can influence the elution pattern of the coumarin mixture. It is obvious that, for this simple mixture, a gradient is not really necessary to resolve all peaks. However, selectivity, resolution, and analysis time can be influenced with a gradient, and the results given in Figure 4 indicate that different gradient conditions are easily established in the microchip system. The gradient traces shown in Figure 4a-c are shifted by the retention time of the nonretained dye C440 measured at the initial gradient conditions. Assuming that the gradient front moves through the channel at the speed of the EOF, these traces now appear as if they were monitored at the detection point close to the end of the separation channel. In this way, it is also easier to see how much of the gradient was “experienced” by each band in the mixture. For example, C440 was not affected by the gradient in any of the experiments shown in Figure 4. This makes sense, since the gradient moves at the same speed as an unretained band and would have to be started before the injection in order to have an effect on C440. C450, on the other hand, is only slightly influenced by the gradient in Figure 4a, whereas there is greater effect apparent from the gradients in Figure 4b and c. By adjusting the gradient parameters start time, start percentage of organic modifier, duration, and slope, one has an effective tool to manipulate the selectivity and optimize the separation for an entire mixture or for a critical peak pair. Another justification for choosing gradient over isocratic elution can be seen when comparing peak widths for peaks which have spent the same amount of time in the chromatographic system. For example, the analysis times for the mixture using 29% acetonitrile isocrati(45) Snyder, L. R.; Kirkland, J. J. Introduction to Modern Liquid Chromatography; 2nd ed.; J. Wiley & Sons: New York, 1979. (46) Balchunas, A. T.; Sepaniak, M. J. Anal. Chem. 1988, 60, 617. (47) Sepaniak, M. J.; Swaile, D. F.; Powell, A. C. J. Chromatogr. 1989, 480, 185. (48) Yan, C.; Dadoo, R.; Zare, R. N.; Rakestraw, D. J.; Anex, D. S. Anal. Chem. 1996, 68, 2726. (49) Huber, C. G.; Choudhary, G.; Horva´th, C. Anal. Chem. 1997, 69, 4429.
Analytical Chemistry, Vol. 70, No. 15, August 1, 1998
3295
Figure 5. Fast open-channel electrochromatography on microchips. Channel depth, 5.2 µm; stationary phase, octadecylsilane; mobile phase, 10 mM borate buffer (pH 8.4) with a linear gradient of acetonitrile from 29% to 50% within 5 s, starting 1 s after injection; analytes, same as in Figure 1; field strength, 700 V/cm; dotted line, gradient trace. Figure 4. Gradient elution in open-channel electrochromatography on microchips. Channel depth, 5.2 µm; stationary phase, octadecylsilane; mobile phase, 10 mM borate buffer (pH 8.4) with different amounts of acetonitrile, (a) linear from 15% to 50% within 15 s, starting 8 s after injection, (b) linear from 15% to 50% within 10 s, starting 5 s after injection, and (c) linear from 29% to 50% within 5 s, starting 3 s after injection; analytes, same as in Figure 1; field strength, 500 V/cm; dotted line, gradient trace.
cally (Figure 1b) and a gradient as in Figure 4c are comparable, but the peak widths are much narrower in the case of the gradient run. At the same time, the signal-to-noise ratio and, hence, the detectability are enhanced. This compression of the bands is explained by the continuous shifting of the partitioning equilibria toward the mobile phase by the gradient front. While molecules which are at the front of the band still experience a less eluotropic mobile phase and spend on average more time in the stationary phase, molecules in the trailing edge of the band are sped up by the following higher eluotropic mobile phase. A similar “mending” effect was seen for tailing peaks in traditional gradient liquid chromatography.45 However, because LC gradients usually have to be much slower than demonstrated in this work in order not to destroy resolution, such “focusing” effects are less pronounced. It must be noted that, in electroosmotically driven solvent programming, such an improvement of peak widths is not automatically seen when running a gradient but occurs only if the electroosmotic flow is not significantly diminished by the increasing eluotropy of the mobile phase counteracting any “focusing” effect. Acetonitrile modifier fulfills this requirement, whereas methanol decreases the EOF more strongly. Indeed, in previously described solvent-programmed MEKC separationss where the retention mechanism is different but comparablesno focusing and, possibly, some extra broadening was observed for later-eluting peaks when methanol gradients were applied.21 3296 Analytical Chemistry, Vol. 70, No. 15, August 1, 1998
In Figure 5, a fast separation of the four-component coumarin mixture in a C18-coated microchip with 5.2-µm-deep channels is shown. The field strength is 700 V/cm and corresponds to a linear velocity of 1.7 mm/s. Although this linear velocity is outside the van Deemter optimum (see Figure 3), in combination with a steep gradient (within 5 s from 29 to 50% acetonitrile, started 1 s after injection), excellent peak shapes and resolution are achieved. All four analyte peaks elute within a 6-s window, and the entire chromatogram is completed in just under 20 s. Including a reconditioning step, the chip is ready for another analysis in ∼60 s. An interesting issue is whether the gradient traces plotted in Figures 4 and 5 mirror exactly the actual gradient inside the channel. The precise control over the mixing of two fluids has been shown for a pure buffer solution and one doped with a fluorescent dye.21 If the solutions to be mixed differ distinctly in their properties (namely conductivity, dielectric constant, and viscosity), the situation is changed, however. Control of the mixing ratio at the mixing tee depends greatly on the electroosmotic flow, which, in turn, is governed by the composition of the mobile phase. Conductivity changes in one leg of the chip will affect current flow and voltages and, hence, electroosmotic flow throughout the microchip. As an overall effect, it is to be expected that the de facto gradient deviates more or less from its intended form and slope. The magnitude of this deviation is dependent on the difference in the conductivities and the electroosmotic flows in the two mixing channels. This deviation is not due to a strongly delayed delivery of the mixed solvents to the head of the separation channel or the large volume of a mixing chamber, because the monolithic design of microchips results in virtually dead-volume-free connections and extremely short transfer tubes. From a purely practical point of view, therefore, this deviation
may be considered an academic problem which may almost be neglected, provided it is reproducible. Very rarely will it be possible to define a gradient a priori which will be the most suitable for a separation problem. More often, the best separation parameters (e.g., the best gradient form, slope, and initial and final conditions) will have to be established empirically with or without the help of optimization algorithms. In these cases, the speed and the ease with which gradients are controlled and executed on-chip, together with short analysis times, allow one to arrive at optimum separation conditions within a few iterative steps, even without knowing the true gradient behavior. In conclusion, the combination of microchip open-channel electrochromatography and on-chip-generated solvent programming displays the potential of being a powerful and versatile analytical tool. The achievable efficiencies and, especially, the fast analyses times compare very well to those of conventional capillary setups. Due to the monolithic design of the microchip, fluidic control is facilitated enormously over capillary networks. In combination with computer control, this approach becomes very conducive to fast optimization of chromatographic separations. By the same token, the integration into automated systems with
artificial intelligence feedback should also be feasible. In the future, other coating procedures and ways to generate stationary phases will be investigated, which should help to increase the capacity of the chromatographic system and avoid overloading issues. Also, the potential possibilities of reversed-phase coated channels for on-chip solid-phase enrichment procedures will be explored. ACKNOWLEDGMENT Oak Ridge National Laboratory is managed by Lockheed Martin Energy Research Corporation for the U.S. Department of Energy, under Contract No. DE-ACO5-96OR22464. This research was supported in part by an appointment for J.P.K. to the Oak Ridge National Laboratory Postdoctoral Research Associates Program, administered jointly by the Oak Ridge National Laboratory and the Oak Ridge Institute for Science and Education. The authors thank Justin E. Daler for preparation of the microchips. Received for review December 19, 1997. Accepted May 7, 1998. AC971367Y
Analytical Chemistry, Vol. 70, No. 15, August 1, 1998
3297
