Use of Folded Micromachined Pillar Array Column with Low

Jan 21, 2010 - the bird's-eye view and cross-sectional view of the sample injection zone, respectively. A deeper channel was fabricated for the sample...
1 downloads 0 Views 3MB Size
Download PDF
Anal. Chem. 2010, 82, 1420–1426

Use of Folded Micromachined Pillar Array Column with Low-Dispersion Turns for Pressure-Driven Liquid Chromatography Chiaki Aoyama,† Akira Saeki,‡ Masao Noguchi,‡ Yoshitaka Shirasaki,§ Shuichi Shoji,‡ Takashi Funatsu,† Jun Mizuno,*,‡ and Makoto Tsunoda*,† Graduate School of Pharmaceutical Sciences, The University of Tokyo, Tokyo, Japan, Major in Nano-Science and Nano-Engineering, Waseda University, Tokyo, Japan, and Kazusa DNA Research Institute, Chiba, Japan In this study, we show for the first time that the separation efficiency of a pillar array column under pressure-driven liquid chromatography (LC) conditions can be improved using a separation channel with low-dispersion turns. The pillar array column was fabricated by reactive ion etching of a silicon substrate. With the low-dispersion-turn geometry, a column with a length and width of 110 mm and 400 µm, respectively, could be fabricated on a 20 × 20 mm microchip. Under nonretained conditions, the solute bands obtained for fluorescent compounds remained almost unchanged even after passing through the lowdispersion turns; however, significant skewing of the solute bands was observed in the case of constant-radius turns. Two coumarin dyes were well resolved under reversed-phase conditions, and a maximum theoretical plate number of 8000 was obtained. Successful separation of the fluorescent derivatives of six amino acids was achieved in 140 s. These results indicated that the separation efficiency of microchip chromatography could be significantly improved using a long separation channel with low-dispersion turns. Pressure-driven high-performance liquid chromatography (HPLC) is a powerful separation technique with high reliability and a versatile pattern of separation. Hence, HPLC plays an important role in various fields such as life science and clinical research, as well as in the food industry. HPLC columns packed with micrometer-sized spherical particles have been widely used since the late 1970s. Despite major research efforts and improvements in column packing bead synthesis, Knox found that even the best possible packed columns do not surpass the barrier of a minimal reduced plate height (h) of around 2.1,2 He found that the major factor that contributes to band broadening is the irregular structure of the mobile phase zone. This finding suggests that the separation efficiency can be * To whom correspondence should be addressed. E-mail: [email protected] (M.T.), [email protected] (J.M.). Phone: +81-3-5841-4761 (M.T.), +81-3-3205-3181 (J.M.). Fax: +81-3-5802-3339 (M.T.), +81-3-3205-3182 (J.M.). † The University of Tokyo. ‡ Waseda University. § Kazusa DNA Research Institute. (1) Knox, J. H. J. Chromatogr., A 1999, 831, 3–15. (2) Knox, J. H. J. Chromatogr., A 2002, 960, 7–18.

1420

Analytical Chemistry, Vol. 82, No. 4, February 15, 2010

significantly improved using columns with a perfectly ordered internal structure. Inspired by Knox’s study, Regnier et al. developed a micromachined pillar array column.3-6 Using semiconductor device fabrication, a high-precision technique, they could control the position of each individual pillar perfectly. This approach is expected to help eliminate the major sources of irregularity in the mobile phase zone. Regnier et al., however, focused on electrodriven separation and not on pressure-driven separation. With the advent of new, highly accurate etching techniques such as the Bosch process, etching of micropillar arrays with very high spatial resolution on a Si wafer has become possible.7,8 By making use of these new machining advantages, De Malsche et al. carried out reversed-phase separation in pillar array columns under pressure-driven liquid chromatography (LC) conditions.9 They achieved h values as low as 1, which is close to the theoretically calculated minimum value. However, since De Malsche et al. used short columns with a length of 10 mm, the maximum theoretical plate number they could obtain was 4000-5000. For practical purposes, it is desirable to use longer pillar array columns in order to increase the theoretical plate number and, thus, achieve high-resolution separation. Since the chip area is small, pillar array columns with turns must be fabricated so that a sufficiently long column can be accommodated on the chip. However, the turns in the separation column generally cause tremendous skewing of the flat solute bands because of the locally nonuniform fluid velocity and, thus, deteriorate the separation performance. Therefore, it is difficult to design a long pillar array column with a low plate height. In the present study, a pillar array column with optimized turn geometry for chip electrophoresis was used. Griffiths et al. designed “low-dispersion turns” using computational fluid dynamics.10 The geometry features an outer boundary with a fixed radius He, B.; Regnier, F. J. Pharm. Biomed. Anal. 1998, 17, 925–932. He, B.; Tait, N.; Regnier, F. Anal. Chem. 1998, 70, 3790–3797. Regnier, F. E. J. High Resolut. Chromatogr. 2000, 23, 19–26. Slentz, B. E.; Penner, N. A.; Regnier, F. J. Sep. Sci. 2002, 25, 1011–1018. Park, W. J.; Kim, J. H.; Cho, S. M.; Yoon, S. G.; Suh, S. J.; Yoon, D. H. Surf. Coat. Technol. 2003, 171, 290–295. (8) De Pra, M.; Kok, W. T.; Gardeniers, J. G. E.; Desmet, G.; Eeltink, S.; van Nieuwkasteele, J. W.; Schoenmakers, P. J. Anal. Chem. 2006, 78, 6519– 6525. (9) De Malsche, W.; Eghbali, H.; Clicq, D.; Vangelooven, J.; Gardeniers, H.; Desmet, G. Anal. Chem. 2007, 79, 5915–5926. (10) Griffiths, S. K.; Nilson, R. H. Anal. Chem. 2001, 73, 272–278. (3) (4) (5) (6) (7)

10.1021/ac902491x  2010 American Chemical Society Published on Web 01/21/2010

and an inner boundary that gradually tapers toward the outer boundary prior to each turn. In this design, the distance traveled by the solute molecules along the inner and outer paths is the same, and therefore, the turn-induced broadening of a solute band is minimized. “Low-dispersion turns” have been used for effective electrophoretic separation.11-13 Although Griffiths et al. mentioned that channels with the above-mentioned geometry are applicable to pressure-driven flows, no experimental studies have been conducted to verify if this geometry is suitable for pressure-driven separation. Therefore, we fabricated a pillar array structure in a long channel with low-dispersion turns and carried out pressuredriven reversed-phase chromatographic separation in a column with a length of at least 110 mm. In this study, we report for the first time that low-dispersion turns are suitable for pressure-driven flows in a pillar array channel. EXPERIMENTAL SECTION Fabrication of Microchips with Pillar Array Columns. The microchips were fabricated by multistep ultraviolet (UV) photolithography and deep reactive ion etching (RIE) using the Bosch process. First, a 250 nm thick silicon oxide layer was formed on a 20 × 20 mm silicon substrate by thermal oxidation at 1100 °C. The pillar array pattern was defined by UV photolithography (MA6, SUSS MicroTec, Germany) in a masking layer comprising a positive photoresist (TSMR-V90, Tokyo Ohka Kogyo, Kanagawa, Japan). The mask pattern was transferred to the underlying silicon oxide layer by wet etching using buffered hydrofluoric acid (BHF 110U, Daikin Industries, Osaka, Japan). A second lithography step was carried out to fabricate an injection channel. A positive photoresist was spin-coated on the etched silicon oxide layer and exposed to UV light. After the exposed silicon oxide layer was wet-etched, the underlying silicon layer was dry-etched by deep RIE (Multiplex-ICP, Sumitomo Precision Products, Hyogo, Japan) to fabricate a 30 µm deep injection channel. After resist stripping, another 30 µm deep channel was dry-etched on the silicon layer by deep RIE to form the pillar array channel and the injection channel. The depth of the pillar channel was 30 µm, and the total depth of the injection channel was 60 µm. Through holes for inlets and outlets were fabricated by backside photolithography and deep RIE. The surface of the microchannel was oxygenated using oxygen plasma (RIE-10NR, Samco, Kyoto, Japan) for the subsequent surface modification process. Finally, the fabricated silicon substrate was bonded to a Pyrex glass plate (Corning 7740, Corning Inc., New York, NY) by anodic bonding (NEC Corporation, Tokyo, Japan) to seal the fluid channels. Chip Coating. For reversed-phase separation, the surface of the separation channel was coated with a monolayer of hydrophobic octadecylsilyl groups by covalent bonding. Since it was necessary to activate the channel surface before the reaction, the following solutions were passed through the pillar array column with the help of a micro-HPLC pump, MP711 (GL Sciences, Tokyo, Japan): water for 30 min, 0.2 M aqueous sodium hydroxide (11) Molho, J. I.; Herr, A. E.; Mosier, B. P.; Santiago, J. G.; Kenny, T. W.; Brennen, R. A.; Gordon, G. B.; Mohammadi, B. Anal. Chem. 2001, 73, 1350–1360. (12) Ramsey, J. D.; Jacobson, S. C.; Culbertson, C. T.; Ramsey, J. M. Anal. Chem. 2003, 75, 3758–3764. (13) Mellors, J. S.; Gorbounov, V.; Ramsey, R. S.; Ramsey, J. M. Anal. Chem. 2008, 80, 6881–6887.

solution/ethanol (40/60, v/v) for 120 min, water for 30 min, 0.1 M hydrochloric acid for 30 min, and ethanol for 30 min. The flow rate was maintained at 0.5 µL/min. The tubing used to connect the pump and the microchip comprised fused silica capillaries with an I.D. of 30 µm. The connection between the pump and the microchip was established by a custom-made setup, in which the chip was sandwiched between two plates of a holder. The top plate was made of polyetheretherketone (PEEK) and perforated with screw holes to fix the capillaries using a nut and ferrule over a 1/16 in. tubing sleeve. The top plate and the chip were strongly sealed with an O-ring made of a perfluoroelastomer to prevent leakage at high pressures. The bottom plate was made of stainless steel with the central part removed for observing the microchip channels with a microscope during the separation experiments. Then, the microchip was dried overnight at 110 °C in an SSSF111 constant-temperature oven (Isuzu Seisakusho, Tokyo, Japan). Dimethyloctadecylchlorosilane (Tokyo Chemical Industry, Tokyo, Japan) was dissolved in dehydrated toluene (Kanto Chemical, Tokyo, Japan), and the microchip was soaked in this reaction solution for 24 h at room temperature. Prior to use, the microchip was flushed with toluene, ethyl acetate, and methanol. The solvents were eluted at a flow rate of 0.5 µL/min using the microHPLC pump. Procedures for Experiments Using the Microchips. Coumarin dyes C525 and C545 were purchased from Exciton (Dayton, OH). These fluorescent dyes were chosen because they have high quantum yields even under aqueous conditions.14 The dyes were dissolved in acetonitrile, and the solution was delivered to the pillar array microchip using a commercially available syringe. Sample injection was carried out by opening the sample inlet and outlet, which were closed during separation. The sample flow was controlled with an external four-port valve (Valco Instruments, Houston, TX). For the experiments under nonretained conditions, acetonitrile was used as the mobile phase at a flow rate of 1 µL/min. The injected sample was a 0.2 mM C525 solution. A fluorescent sample band was observed using an IX70 inverted microscope (Olympus, Tokyo, Japan). Excitation was carried out using a U-PS50MH halogen lamp (Olympus). The filter cube consisted of a BP460-490 excitation filter (Olympus), a 505DRLP dichroic mirror (Omega Optical, Brattleboro, VT), and an HQ535m emission filter (Chroma Technology, Rockingham, VT). A UPlanFL 4× (N.A. 0.13, PhL, Olympus) objective and a 330RCX CCD camera (Dage-MTI, Michigan City, IN) mounted on a U-TV0.35XC-2 video adaptor (Olympus) were used to observe the geometry of each turn. For the experiments on coumarin dyes under reversed-phase LC conditions, a water/acetonitrile (65/35, v/v) mixture was used as the mobile phase. A solution of C525 (0.2 mM) and C545 (0.4 mM) was injected into the column. The microscope system was suitably modified to achieve high detection sensitivity and good linearity. The UPlanFL 4× objective was replaced with a UPlanApo 10× (N.A. 0.40, Olympus) objective, and the fluorescence intensity was analyzed using an H7421-40 photomultiplier tube (Hamamatsu Photonics, Hamamatsu, Japan) equipped with a PHC-2500 photocounter (Scientex, Hamamatsu, Japan). A PSL-3F slit (Sigma Koki, Tokyo, Japan) was placed between the emission filter and (14) Fletcher, A. N.; Bliss, D. E.; Kauffman, J. M. Opt. Commun. 1983, 47, 57–61.

Analytical Chemistry, Vol. 82, No. 4, February 15, 2010

1421

Figure 1. Overview of fabricated microchip. Pillar array column with (A) low-dispersion turns and (B) constant-radius turns.

the photomultiplier tube to confine the detection area. The theoretical plate number (N) and plate height (H) were obtained from the following equations: N ) 5.54tR2 /w1/22 H ) L/N where tR is the retention time, w1/2 is the peak width at half peak height, and L is the column length. The total length of the turns is defined as the sum of the outer radii of the turns or as the average of the inner and outer radii of the turns for low-dispersion turns and constant-radius turns, respectively. For the separation of amino acids, a water/acetonitrile/ trifluoroacetic acid (90/10/0.02, v/v/v) mixture was used as the mobile phase. The flow rate was set at 0.5 µL/min. 4-Fluoro-7nitro-2,1,3-benzoxadiazole15 (NBD-F, Dojindo Laboratories, Kumamoto, Japan) was used as a fluorescent derivatization reagent for amino acids. The derivatization procedures were based on those reported in our previous studies.16-18 To 40 µL of a standard solution containing six amino acids (0.5 mM each of ε-amino-ncaproic acid, isoleucine, leucine, phenylalanine, proline, and valine) were added 30 µL of 0.2 M borate buffer (pH 8.5) and 30 µL of 10 mM NBD-F in acetonitrile. After heating the mixture in a water bath at 60 °C for 3 min, the reaction was stopped by adding 100 µL of 1 M tartrate buffer (pH 2.0). Since the fluorescence intensity of the NBD derivatives was not as strong as that of the coumarin dyes, excitation was carried out by irradiation with a solid-state blue laser (Sapphire 488-20, Coherent Japan, Tokyo, Japan). The BP460-490 excitation filter was removed to prevent the generation of interference fringes. The UPlanApo 20× (N.A. 0.70, Olympus) objective was used for more effective fluorescence correction. Detection was carried out with an H7421-40 photomultiplier tube. (15) Watanabe, Y.; Imai, K. Anal. Biochem. 1981, 116, 471–472. (16) Aoyama, C.; Santa, T.; Tsunoda, M.; Fukushima, T.; Kitada, C.; Imai, K. Biomed. Chromatogr. 2004, 18, 630–636. (17) Aoyama, C.; Tsunoda, M.; Funatsu, T. Anal. Sci. 2009, 25, 63–65. (18) Nonaka, S.; Tsunoda, M.; Imai, K.; Funatsu, T. J. Chromatogr., A 2005, 1066, 41–45.

1422

Analytical Chemistry, Vol. 82, No. 4, February 15, 2010

RESULTS AND DISCUSSION Design of Microchip with a Folded Pillar Array Column Having Low-Dispersion Turns. In this study, to achieve high separation efficiency, a long pillar array column with lowdispersion turns was fabricated on a microchip. Low-dispersion turns were designed on the basis of the simulation results obtained by Griffiths et al.10 The width of the low-dispersion turns (110 µm) was approximately 28% of that of the straight channel (400 µm). The inner boundary gradually tapered toward the outer boundary approximately two channel widths before each turn. Figure 1 shows the overview of the two types of microchips fabricated with pillar array columns: one with straight channels and low-dispersion turns (Figure 1A) and the other with straight channels and constant-radius turns (Figure 1B). The latter was used for comparison. Because of the formation of turns, a 110 mm long separation channel (with 12 turns) could be accommodated on both the microchips. Both the chips included an injection channel as well as a separation channel. Figure 2 shows the scanning electron microscopy (SEM) images of the microchips with folded pillar array columns. It is to be noted that pillar arrays were formed both in the turns and in the straight channels. The pillar array consisted of square pillars (side: 3 µm) separated by a distance of 2 µm (Figure 2A). On the photolithographic mask, 4 µm square pillars were formed at 1 µm intervals. These differences between the experimental and designed values were mainly due to the isotropic wet etching of the silicon oxide layer; the etching process led to the formation of lateral undercuts, which in turn caused pattern broadening. Therefore, the observed external porosity ε was 0.6, while the ε value determined on the basis of the results of a theoretical study was approximately 0.4.19 The geometry of the sidewall plays a critical role in the overall band broadening.8 In a previous study, sidewalls with two different geometries, a flat sidewall and a sidewall with embedded pillars, were experimentally tested, and the latter was found to be better. (19) De Smet, J.; Gzil, P.; Vervoort, N.; Verelst, H.; Baron, G. V.; Desmet, G. J. Chromatogr., A 2005, 1073, 43–51.

Figure 2. SEM images of the fabricated microchip. (A) Enlarged view of pillar array structure and (B) enlarged view of the structure near the inner boundary of the low-dispersion turns. (C) Bird’s-eye view and (D) cross-sectional view of sample injection zone.

Hence, in the present study, sidewalls with embedded pillars are fabricated. The depth of the pillar array column is 30 µm, which is greater than the previously reported value of 10-15 µm.9 This deep column has two advantages: (1) suppression of the band broadening effect caused by the top and bottom walls;20 and (2) stable flow caused by an increase in the flow rate at a constant linear flow velocity. Figure 2B shows the pillar array structure near the channel boundary of the bent region. The boundaries of the low-dispersion turns and constant-radius turns were shaped on the basis of the original geometry by maintaining the uniformity of the pillar array structure to the maximum possible extent. Figure 2C,D shows the bird’s-eye view and cross-sectional view of the sample injection zone, respectively. A deeper channel was fabricated for the sample solution. In the sample introduction step, the sample solution was allowed to flow temporarily through the channel so that the abovementioned deeper channel was filled with the sample. Subsequently, the sample flow was stopped, and the sample in the mobile phase stream was carried to the pillar array column. There is a strong demand for high-pressure-resistant couplings that can be used in lab-on-a-chip devices.21 In this study, a custommade holder was used for the chip, and no leakage was observed up to pressures of 70 bar. Under normal separation conditions (mobile phase: water/acetonitrile (65/35, v/v); flow rate: 1 µL/ min), the pressure was 18 and 11 bar for the chip with lowdispersion turns and constant-radius turns, respectively. This suggested that the current setup for pressure-driven LC was suitable for normal separation conditions. The difference between the pressures in the chip with low-dispersion turns and that with constant-radius turns could be ascribed to the fact that the width (20) Eghbali, H.; De Maische, W.; De Smet, J.; Billen, J.; De Pra, M.; Kok, W. T.; Schoenmakers, P. J.; Gardeniers, H.; Desmet, G. J. Sep. Sci. 2007, 30, 2605–2613. (21) Ocvirk, G.; Verpoorte, E.; Manz, A.; Grasserbauer, M.; Widmer, H. M. Anal. Meth. Instr. 1995, 2, 74–82.

of the low-dispersion turns was only 28% of the constant-radius turns. Therefore, for the same column length, the pressure in the low-dispersion turns would be four times that in the constantradius turns. Accordingly, the pressure in the chip with lowdispersion turns was twice that in the chip with constant-radius turns, although the turns occupied only 20% of the total column length. Injection and Separation of Coumarin Dyes in the Straight Channel. To investigate the performance of the pillar array column, separation of C525 and C545 was carried out in the straight channel. The injection band was rectangular, indicating that the injection process was successful (Figure 3). In the case of C525, the theoretical plate number calculated at a point 7 mm downstream of the injection zone was 775, which corresponded to an H value of 9.3 µm. De Malsche et al.9 calculated H by measuring the peak variance at two different positions: one very close to the injection zone and the other approximately 10 mm away from this zone. Thus, they succeeded in nullifying the band broadening effect. Hence, in this study, we recalculated H in a similar manner; the obtained value (4.2 µm) was similar to the previously reported value9 (4 µm). Investigation of Low-Dispersion Turns in the Pillar Array Column. First, the behavior of the flat sample bands in the lowdispersion turns and constant-radius turns was examined under nonretained conditions. Microchips without surface modification were used for this purpose. A solution of C525 was used as the sample, and its fluorescence was detected using a CCD camera; the results are shown in Figure 4. Skewing of the sample band was observed in the constant-radius turns but not in the lowdispersion turns. The H values in the straight channel, lowdispersion turns, and constant-radius turns were calculated. To quantify the variance caused by the turns alone, the peak widths before and after each turn were measured. The H values calculated for the straight channel, low-dispersion turns, and constant-radius turns were 2.7, 3.5, and 510 µm, respectively. This showed that Analytical Chemistry, Vol. 82, No. 4, February 15, 2010

1423

Figure 3. Fluorescent images of sample injection zone under reversed-phase LC conditions when the sample solution is injected (A) and images obtained 1 s (B), 2 s (C), and 3 s (D) after the injection. Sample solution contains two fluorescent dyes (C525 and C545).

Figure 4. Movement of flat sample bands through (A) constantradius turns and (B) low-dispersion turns under nonretained conditions. The top two rows of images show the separation channel filled with the fluorescent solution. Subsequent images were recorded every 0.5 s. 1424

Analytical Chemistry, Vol. 82, No. 4, February 15, 2010

there was no notable contribution to band broadening from the low-dispersion turns and that the turn-induced variance for the low-dispersion turns was 150 times lower than the variance for the constant-radius turn. This result was consistent with that of Griffiths and Nilson’s theoretical study, according to which the turn-induced variance could be reduced by 2 to 3 orders of magnitude by adopting low-dispersion turns.10 Our experimental results demonstrated for the first time that low-dispersion turns are suitable for the pillar array channel. Second, the effect of low-dispersion turns on the separation efficiency was examined under reversed-phase LC conditions. The separation of C525 and C545 was carried out using a mixture of water/acetonitrile (65/35, v/v) as the mobile phase. As shown in Figure 5, two peaks with good baseline separation could be detected at points 7.0, 41.2, 75.4, and 107 mm downstream of the injection zone. The peak resolution improved with an increase in the number of turns through which the sample bands passed. The resolution corresponding to position A in Figure 5 was 4.43, and this value increased to 13.7 at the final detection point (position D in Figure 5). The number of theoretical plates and the plate heights calculated from the chromatograms are shown in Table 1. The number of theoretical plates increased linearly with the distance between the injection zone and the detection point. A similar increase in N was observed for the microchip with the constant-radius turns as well; nevertheless, the N value achieved with the low-dispersion turns was significantly higher than that achieved with the constant-radius turns. The high N values obtained for the low-dispersion turns confirmed the

Figure 5. Chromatograms of C525 and C545 separation achieved on the microchip with low-dispersion turns. Detection was carried out at points (A) 7.0 mm, (B) 41.2 mm, (C) 75.4 mm, and (D) 107 mm downstream of the injection zone. Peaks: 1, C525; 2, C545. Table 1. Comparison of Numbers of Theoretical Plates (N) and Plate Heights (H) Corresponding to the C525 and C545 Peaks at Several Points in the Columna constant-radius turns low-dispersion turns detection points C525 A B C D C545 A B C D a

N

H (µm)

N

H (µm)

694 1498 2425 3017 578 1244 1934 2678

10.1 28.9 32.5 32.9 12.1 31.2 36.5 37.1

755 3628 5616 8156 556 2465 4298 6239

9.3 11.4 13.4 13.1 12.6 16.7 17.5 17.1

Detection points are the same as those shown in Figure 5.

improved separation performance of this microchip. In the case of the microchip fabricated with low-dispersion turns, the N value for a 110 mm long column was 8000, which was almost thrice that obtained with the microchip having constant-radius turns. This result demonstrated that folded long pillar array columns with low-dispersion turns allow improved LC separation. On the basis of the results obtained for three different chips, we found that the reproducibility of the retention times of the coumarin dyes was 7-12%, which was probably due to the fabrication process and chip coating procedure adopted. The N value of 8000 obtained for a column length of 110 mm is not satisfactory, and hence, optimization studies must be carried

out. According to the results of a simulation study, the smaller the interpillar distance and the pillar diameter, the better is the separation efficiency.19 The extent to which the pillar size can be reduced depends on the microchip fabrication method. In contrast to wet etching, lateral undercuts are not formed in anisotropic dry etching, and hence, pattern broadening is not observed. Therefore, anisotropic dry etching is an efficient technique that can be used to reduce the interpillar distance. In addition, UV imprint lithography22,23 can be used to fabricate pillar arrays that are smaller than those fabricated by UV photolithography. When the separation efficiency is improved, the width of the sample bands would reduce before each turn in the column, and significant band broadening would occur in the turns. Hence, lowdispersion turns are expected to be more useful for high-efficiency separation. Separation of Fluorescently Labeled Amino Acids. To further investigate the feasibility of the use of pillar array columns with low-dispersion turns for reversed-phase LC, we carried out the separation of biological compounds. In this study, we chose amino acids that were derivatized with NBD-F, which was investigated in our previous studies.16–18 Figure 6 shows the chromatograms obtained for a mixture of six fluorescently labeled (22) Haisma, J.; Verheijen, M.; van den Heuvel, K.; van den Berg, J. J. Vac. Sci. Technol., B 1996, 14, 4124–4128. (23) Chou, S. Y.; Krauss, P. R.; Renstrom, P. J. Appl. Phys. Lett. 1995, 67, 3114– 3116.

Analytical Chemistry, Vol. 82, No. 4, February 15, 2010

1425

amino acids including the byproduct NBD-OH. Only four peaks were detected in the chromatogram just before the sample entered the first turn (Figure 6A); this indicated that the straight channel was not suitable for practical applications and that a longer channel was necessary. As shown in Figure 6B, the separation of the NBDamino acids was improved when the sample was at a point before the third turn. When the sample was at a point before the fifth turn in the column, almost complete baseline separation of seven species was achieved within 140 s (Figure 6C). Because NBD-Ile and NBD-Leu have the same molecular weight, it is difficult to separate them on conventional LC columns. With our method, the resolution factor increased from 0.21 (Figure 6A) to 0.76 (Figure 6B) and then to 0.98 (Figure 6C) with an increase in the column length. Consequently, the separation of NBD-amino acids suggested that a folded separation channel with low-dispersion turns is highly useful for the analysis of complex sample mixtures. CONCLUSIONS Low-dispersion turns were successfully formed in a micromachined pillar array column. Under nonretained conditions, the turninduced variance for low-dispersion turns was 150 times lesser than that for constant radius turns, as proposed by the results of a previous theoretical study. Under reversed-phase LC conditions, two coumarin dyes could be well separated with a resolution of 13.7. Using a long separation channel with low-dispersion turns, the separation efficiency of microchip chromatography could be significantly enhanced, as demonstrated by the separation of the fluorescent derivatives of six amino acids in 140 s. The maximum N value obtained for a 110 mm long column under reversed-phase LC conditions was 8000. Although further optimization of the proposed method might be necessary, the results of the present study show that, when low-dispersion turns are used, high separation efficiency can be achieved in microchip chromatography. With the proposed approach, it is possible to analyze complex biological samples containing various compounds.

Figure 6. Chromatograms obtained from fluorescently labeled amino acid solution on the microchip with low-dispersion turns. Detection was carried out at various points just before the (A) first turn, (B) third turn, and (C) fifth turn. Peaks: 1, NBD-OH; 2, NBDPro; 3, NBD-Val; 4, NBD-ε-amino-n-caproic acid; 5, NBD-Ile; 6, NBD-Leu; 7, NBD-Phe.

1426

Analytical Chemistry, Vol. 82, No. 4, February 15, 2010

ACKNOWLEDGMENT C.A. and A.S. contributed equally to this work. Received for review November 2, 2009. Accepted January 7, 2010. AC902491X