Electroosmotic Flow-Based Pump for Liquid Chromatography on a

Anal. Chem. 2008, 80, 8287–8292

Electroosmotic Flow-Based Pump for Liquid Chromatography on a Planar Microchip Joseph F. Borowsky,† Braden C. Giordano,‡ Qin Lu,† Alex Terray,† and Greg E. Collins*,† Chemistry Division, Naval Research Laboratory, 4555 Overlook Avenue, SW, Code 6112, Washington, D.C. 20375-5342, and Nova Research, Inc., 1900 Elkin Street, Suite 230, Alexandria, Virginia 22308 An electroosmotic flow (EOF)-based pump, integrated with a sol-gel stationary phase located in the electric fieldfree region of a microchip, enabled the separation of six nitroaromatic and nitramine explosives and their degradation products via liquid chromatography (LC). The integrated pump and LC system were fabricated within a single quartz substrate. The pump region consisted of a straight channel (3.0 cm × 230 µm × 100 µm) packed with 5-µm porous silica beads. The sol-gel stationary phase was derived from a precursor mixture of methyltrimethoxy- and phenethyltrimethoxysilanes and was synthesized in the downstream, field-free region of the microchip, resulting in a stationary-phase monolith with dimensions of 2.6 cm × 230 µm × 100 µm. Fluid dynamic design considerations are discussed, especially as they relate to integrating the EOF pump with the LC system. Pump and separation performance, as characterized by flow rate measurements, injection, elution, separation, and detection, point to a viable analytical chemistry platform that encompasses all of the benefits expected of portable, laboratory-on-chip systems, including reduced sample requirements and small packaging. The separation and identification of analytes continues to be an important analytical chemistry procedure with many of the recent advances in the broad field of separation science being attributed to developments in microfluidics. The small length scales associated with microfluidics enable reduced analysis times, smaller sample requirements, and often superior performance in terms of resolution and sensitivity. Separations on the microscale have generally focused on electrophoretic-based processes,1 which rely on differences in electrophoretic mobilities of the analytes and electrolytes under an applied electric field, and liquid chromatography,2,3 which relies on differences in the affinity of the analytes to the mobile and stationary phases. High-performance liquid chromatography (HPLC) has long been a proven separation technique because of excellent performance, good reproducibility, and flexibility in matching the analyte * To whom correspondence should be addressed. E-mail: greg.collins@ nrl.navy.mil. † Naval Research Laboratory. ‡ Nova Research, Inc. (1) Landers, J. P., Ed. Handbook of Capillary and Microchip Electrophoresis and Associated Microtechniques, 3rd ed.; CRC Press: New York, 2008. (2) Karlsson, K. E.; Novotny, M. Anal. Chem. 1988, 60, 1662–1665. (3) Kennedy, R. T.; Jorgenson, J. W. Anal. Chem. 1989, 61, 1128–1135. 10.1021/ac801497r CCC: $40.75  2008 American Chemical Society Published on Web 10/07/2008

set with the appropriate stationary phase.4 To achieve this performance, high pressures are necessary in order to rapidly transport the mobile phase through the small-diameter chromatograpic packings that comprise a separation column. Generally, high pressures have been achieved with mechanical pumps, but the fairly recent development of flow-based electroosmotic or electrokinetic pumping has provided another motive source; application of an electric field through small pores or channels results in flow rates at high pressure in downstream, electric fieldfree regions.5,6 Early electrokinetic pumping research focused on capillary platforms, and not surprisingly, some of the first applications of electrokinetic pumping dealt with capillary-based liquid chromatography.7,8 The use of the electrokinetic pumping on microchips for separations has not been extensively reported due to the difficulty in achieving sufficient pressures and flow rates to overcome the back pressure of stationary phases located in the field-free region. Nonetheless, recognizing the benefits of coupling the robust performance of LC with the portability and ultimate ease in fabrication and integration of various fluidic components, the application of LC to microchips has been considered.9-15 These applications often relied on external valves or external syringes as a pressure-driven motive source. Lazar and co-workers presented a truly integrated LC device for proteomic analysis using unpacked, multichannel electroosmotic flow (EOF) pumping as the fluid motive force, an on-chip C-18 bead bed for separation, and an electrospray ionization (ESI) interface for coupling to off(4) Stevenson, R.; Johnson, E. L. Basic Liquid Chromatography; Varian Associates: Palo Alto, 1978. (5) Arnold, D. W.; Neyer, D. W.; Paul, P. H. Abstr. Paper Am. Chem. Soc. 2000, 219, U97. (6) Zeng, S. L.; Chen, C. H.; Mikkelsen, J. C.; Santiago, J. G. Sens. Actuators, B 2001, 79, 107–114. (7) Chen, L. X.; Ma, J. P.; Guan, Y. F. J. Chromatogr., A 2004, 1028, 219–226. (8) Wang, L.; He, Y. Z.; Deng, N.; Wang, X. K.; Fu, G. N. Instrum. Sci. Technol. 2006, 34, 743–753. (9) Ericson, C.; Holm, J.; Ericson, T.; Hjerten, S. Anal. Chem. 2000, 72, 81– 87. (10) Ishida, A.; Yoshikawa, T.; Natsume, M.; Kamidate, T. J. Chromatogr., A 2006, 1132, 90–98. (11) Murrihy, J. P.; Breadmore, M. C.; Tan, A. M.; McEnery, M.; Alderman, J.; O’Mathuna, C.; O’Neill, A. P.; O’Brien, P.; Advoldvic, N.; Haddad, P. R.; Glennon, J. D. J. Chromatogr., A 2001, 924, 233–238. (12) Ocvirk, G.; Verpoorte, E.; Manz, A.; Grasserbauer, M.; Widmer, H. M. Anal. Methods Instrum. 1995, 2, 74–82. (13) Penrose, A.; Myers, P.; Bartle, K.; McCrossen, S. Analyst 2004, 129, 704– 709. (14) Reichmuth, D. S.; Shepodd, T. J.; Kirby, B. J. Anal. Chem. 2005, 77, 2997– 3000. (15) Yin, N. F.; Killeen, K.; Brennen, R.; Sobek, D.; Werlich, M.; van de Goor, T. V. Anal. Chem. 2005, 77, 527–533.

Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

8287

chip mass spectrometry.16 While the separation was highly efficient (15-30 s peak widths at half-height), the separation time was long (40 min). Alternatively, electrolysis-based pumps have been integrated into microfluidic devices for LC applications.17,18 Electrolysis-based pumps rely on the generation of air bubbles in a sealed reservoir that is incorporated on the microchip device. The bubbles, formed by electrolysis, displace the liquid in the reservoir into the surrounding fluidic network. Efforts by Lee et al. focused on the development of an LC device for the analysis of peptide mixtures.18 Their device included three electrochemical pumps, one for sample injection and the other two for the generation of a solvent gradient coupled to a C-18 packed bead bed with an ESI interface for coupling to off-chip tandem-mass spectrometry. Similar to Lazar and co-workers,16 efficiency was high, but separation times were long (∼10-30 s peak widths at half-height, ∼45 min total separation time). Fuentes and Woolley presented the integration of an electrolysis-based pump with an open-tubular LC separation of three FITC-labeled amino acids.17 Separations were very fast, finishing in less than 40 s, with efficiencies (N) of >3000 plates (or 120 000 plates/m). The present work focuses on the results of an integrated LC separation system on a planar microchip platform. Nitroaromatic and nitramine explosives and their degradation products were used as an analyte test set due to broad Department of Defense interest in developing new, rapid analysis techniques for explosives detection for both public safety and military applications as it relates to improvised explosives detection (IED) and environmental protection, for example. Because explosives are small molecules lacking any fluorescence, this analyte set also demonstrates the power of our system for obtaining two-dimensional absorbance plots (wavelength as a function of retention time) for micro LC in real time, a detection technique that has much broader applicability. This proof-of-concept design is based on modifications to a microchip EOF-based pump previously demonstrated by our group to achieve high pressures on a planar chip platform.19 The integration of the EOF pump, injection system, LC-based separation, and on-chip detection is discussed in terms of microchip design as well as pump and LC system performance. EXPERIMENTAL SECTION Reagents. Ammonium hydroxide, hydrofluoric acid (49%), sodium hydroxide, and acetonitrile (HPLC grade) were purchased from Fischer Scientific (Fair Lawn, NJ). Nitric acid and hydrogen peroxide were purchased from Acros (Fair Lawn, NJ). Acetic acid, sodium silicate solution (∼14% NaOH, 27% SiO2), isopropyl alcohol, poly(ethylene glycol) (PEG), methyltrimethoxysilane, phenethyltrimethoxysilane, and 2-amino-2-(hydroxymethyl)-1,3-propanediol (Tris) were purchased from Sigma-Aldrich (St. Louis, MO). Chromium etchant and buffered oxide (6:1) were obtained from Transene Co. (Danvers, MA). Porous silica particles (10 and 5 µm in size, 50-µm pore) were obtained from Macherey-Nagel (16) Lazar, I. M.; Trisiripisal, P.; Sarvaiya, H. A. Anal. Chem. 2006, 78, 5513– 5524. (17) Fuentes, H. V.; Woolley, A. T. Lab Chip 2007, 7, 1524–1531. (18) Xie, J.; Miao, Y. N.; Shih, J.; Tai, Y. C.; Lee, T. D. Anal. Chem. 2005, 77, 6947–6953. (19) Borowsky, J.; Lu, Q.; Collins, G. E. Sens. Actuators, B 2008, 131, 333– 339.

8288


Figure 1. Diagram of liquid chromatography microfluidic device including closeups of the weir structure and the sol-gel stationary phase. Units are in millimeters. PI ) pump inlet, PO ) pump outlet, SI ) sample inlet, and W ) waste.

(Bethlehem, PA). Individual explosives standards, including RDX, 1,3,5-trinitrobenzene (1,3,5-TNB), nitrobenzene (NB), 2,4,6-trinitrotoluene (TNT), 2-amino-4,6-dinitrotoluene (2-Am-4,6-DNT), and 4-nitrotoluene (4-NT), were purchased from Supelco (Bellefonte, PA) at a concentration of 1000 µg/mL in acetonitrile. Deionized water was obtained from a Milli-Q Plus (Millipore, Billerica, MA) water system. Microfluidic Design. The EOF pump, consisting of a microcolumn packed with porous silica particles, and the LC sol-gel monolith were integrated on a single quartz microchip (Figure 1). This proof-of-concept design is an extension of a high-pressure, EOF pump tested previously,19 now incorporating an addition inlet for sample injection and a downstream field-free region to enable LC separations. The EOF pump region consisted of a 3.0-cm-long straight channel, with the first 0.2 cm of length being packed with 10-µm porous silica particles, followed by 2.8 cm of 5-µm porous silica particles. The front end of the packed column was immobilized behind a weir structure with an opening of ∼8 µm. The silica particles are an ideal packing material for the EOF-based pump, as the packing results in numerous, small interstitial pores with high ζ potential propertiesstwo characteristics necessary to generate high pressure and flow rate. The length of the mechanical weir, ∼450 µm, was minimized in order to reduce the inherent restriction of the shallow region and increase pump flow rate. In order to more closely match the cross-sectional area of the deep, packed bed channel and the shallow weir region, a gradual transition in channel width from 230 to 1200 µm was etched, thereby, mitigating the effects of the flow restriction. A sample inlet (SI) was included between the pump region and the stationary phase. Whereas the previous design had a shallow fieldfree region,19 the present design included a deeper downstream channel (100 µm) in order to accommodate a sol-gel stationary phase and the need for an extended optical path length to enhance sensitivity for absorbance detection. Nominal channel dimensions are given in Table 1.

Table 1. Device Dimensions length (cm) pump region weir pump outlet and sample inlet side arms stationary phase

width (µm)

depth (µm)

3.0 0.045 1.1

230 1200 230

100 7.3 100

3 (2.6 effective)

230

100

Microchip Fabrication. The microchip device was fabricated by standard photolithography and wet etching techniques, using a quartz substrate coated with chromium and photoresist (Nanofilm). Since the device contains a weir structure, two separate exposures were required. The photomask design of the microchip device was transferred to a quartz substrate using a 100-W, 365nm ultraviolet flood lamp (SB-100PC; Spectroline). The exposed photoresist and underlying chrome were removed by immersion in 0.1 N NaOH, followed by chromium etchant. After the first exposure, isotropic etching of the developed substrate using buffered oxide (6:1) formed the microchannels with a channel depth of ∼100 µm. Due to the lengthy etching process required to form the microchannels (∼10 h) on a quartz substrate, the photoresist remaining on the substrate after etching the microchannels was no longer suitable for a second exposure. A new layer of photoresist (Shipley 1813) was spin-coated on the substrate to ensure the quality of the weir formation. The second exposure and subsequent etching with buffered oxide solution provided the weir with a depth of ∼8 µm. After the formation of microchannels and the weir structure, but before the complete removal of the photoresist and chromium coatings, access holes were drilled at the end of each channel using a drill press equipped with a diamond drill bit (Tripple Ripple, Crystalite Corp., Lewis Center, OH). Prior to bonding, the quartz substrate and a quartz cover plate were sequentially cleaned for 30 min in 30% Branson solution, basic piranha solution (1 part hydrogen peroxide, 2 parts concentrated ammonium hydroxide, and 2 parts water), and deionized water. The bond side of the clean, dry quartz substrate and the quartz cover plate were covered with a thin layer of diluted sodium silicate solution (25 wt %) through a syringe equipped with a 0.25-µm syringe filter. The plates were then pressed together under pressure (∼1.5 tons) in a hydraulic press (Carver Inc., Wabash, IN) for 5 min at room temperature to squeeze out the excess silicate solution between the two quartz plates. The silicate solution in the microchannels was removed by applying vacuum at each of the drilled access holes. DI water was used to rinse the microchannels to ensure the complete removal of silicate solution from the microchannel in order to avoid any clogging of the weir structure. After the cleanup of the microchannels, the device was once more placed in the press under 1.5 tons pressure and was heated to 200 °F and maintained at this temperature for 2 h to achieve a fully bonded quartz microchip device. EOF Pump Column Packing. The EOF pump region was formed by packing 10- and 5-µm porous silica particles behind a weir structure, as described under the Microfluidic Design subsection. Packing of the silica particles into the microchannel upstream of the weir was achieved by using a syringe filled with a suspension of 2 wt % silica particles in water and contained within a hand-held syringe pump (Unimicro Technologies, Pleasanton, CA). The connection between the syringe and the microchannel was achieved using a Nanoport fitting (Upchurch Technologies,

Oak Harber, WA) and a fused-silica capillary. Once the packing reached the end of the microchannel, the Nanoport was connected to a pressure cell (Next Advance, Rensselaer, NY) via a capillary. Water was pushed through the column by applying 800 psi pressure from an argon tank for 1 h to afford a packed column with high homogeneity. The column was then flushed with ethanol and dried with argon using the pressure cell. The free end of the column was sealed in place by wetting the very tip of the column with a sodium silicate solution and allowing it to dry at room temperature in a laminar flow hood. This sealing step provided a column with excellent stability under the back pressure generated by the downstream LC column during the pump operation. Sol-Gel Synthesis. The formulation of the sol-gel was similar to the stationary phase described in previous work and consisted of a porogen, monomer, and catalyst.20,21 The porogen was a mixture of 1 mL of 100 mM acetic acid and 0.2 g/mL PEG (Mw 8000), while the monomer consisted of 0.502 mL of methyltrimethoxysilane and 0.086 mL of phenethyltrimethoxysilane. A volume of 0.397 mL of ethanol (200 proof) was added to the monomer and porogen components, and the entire mixture was stirred for 2 h. The catalyst, consisting of 25 µL of diethylamine and 25 µL of acetonitrile, was added to the mixture and slightly hand-stirred. A small volume was placed in the downstream sidearm reservoir hole and was transported into the stationaryphase channel via capillary action. By applying pressure to the waste reservoir, movement of the sol-gel front was stopped before the end of the channel to provide a region free of sol-gel that would maximize absorbance detection capability at the terminal end of the sol-gel column. After allowing the microchip to sit for 3 h at room temperature, the sol-gel was conditioned by flushing with ethanol using a syringe pump. The porosity of the sol-gel formulation was measured by synthesizing a sol-gel within a 100-µm-i.d. capillary. By performing weight differential measurements of the sol-gel under wet and dry conditions, a porosity of 80% was measured, similar to other values reported in the literature.22 Microchip Operation. The microchip integrated all of the components necessary for LC operations, including pressure generation, stationary-phase conditioning, analyte injection, elution, separation, and detection. All steps required for the implementation of this device for liquid chromatography are summarized in Table 2. Prior to use, the pump outlet (PO) was filled with 50 mM Tris and the reservoir sealed with a modified Upchurch fitting that contained a ground electrode epoxied in place. The pump inlet (PI) was filled with the appropriate eluent and the explosives sample was placed in the sample inlet. A 2-kV potential was applied between the sample inlet and waste reservoirs for 100 s, allowing for sample injection and extraction onto the head of the sol-gel. After injection, the sample solution was replaced with 50 mM Tris and the reservoir sealed. The chromatographic separation commenced when a potential was applied between the pump inlet and the pump outlet. Since the pump outlet and sample inlet were sealed, the electroosmotic(20) Giordano, B. C.; Copper, C. L.; Collins, G. E. Electrophoresis 2006, 27, 778–786. (21) Giordano, B. C.; Terray, A.; Collins, G. E. Electrophoresis 2006, 27, 4295– 4302. (22) Breadmore, M. C.; Shrinivasan, S.; Wolfe, K. A.; Power, M. E.; Ferrance, J. P.; Hosticka, B.; Norris, P. M.; Landers, J. P. Electrophoresis 2002, 23, 3487–3495.


8289

Table 2. Summary of Steps for Device Use step no. 1. seal pump outlet 2. sample injection 3. seal sample inlet 4. separation

pump inlet solution reservoir potential solution reservoir potential solution reservoir potential solution reservoir potential

20 mM Tris, 30-70% open na 20 mM Tris, 30-70% open na 20 mM Tris, 30-70% open na 20 mM Tris, 30-70% open 2000-4500 V

acetonitrile acetonitrile acetonitrile

based flow propagated down the sol-gel stationary phase allowing for a field-free separation of analytes. UV Absorbance Detection. Detection of explosives and their degradation products on the microchip was achieved via UV absorbance and has been described previously.23 Briefly, output from a deuterium lamp was carried via an optical fiber to a focusing beam probe containing a series of lenses that focused the light into the separation channel just beyond the sol-gel stationary phase on the microchip. Transmitted light was collected through a quartz microscope objective into a second UV-grade optical fiber, which carried the light to a miniature CCD spectrometer (Ocean Optics, Dunedin, FL). RESULTS AND DISCUSSION EOF Pump Characterization. An important consideration in designing the microchip for LC was the proper coupling of the hydraulic performance of the EOF pump with the separation performance of the stationary phase. The inverse relation between flow rate generated by an EOF pump and the back pressure of the field-free region mandates that the back pressure, evident from a packed bed stationary phase, not be excessively high in order for the electrokinetic pump to generate a sufficient flow rate to achieve separation.6 For this application, an alkylated sol-gel was considered to be an ideal stationary phase for two reasons: (1) unlike standard HPLC particle-based packings, sol-gels have higher porosity and a resulting lower back pressure that puts significantly lower constraints on the design of the EOF micropump; and (2) the alkylated sol-gel utilized here has been demonstrated to be an ideal stationary phase for separating nitramine and nitroaromatic explosives and their degradation products due to its high surface area features that combine hydrophobic and hydrophilic character for separating polar explosive analytes.21 Previous efforts in our laboratory have focused on the use of alkylated sol-gel precursors for the in situ generation of separation columns for microchip-based electrochromatography.21 We demonstrated that inclusion of methyl- and ethyltrimethoxysilanes in the sol-gel solution results in a stationary phase with sufficient hydrophobicity to separate several nitroaromatic and nitramine explosives and their degradation products. Taking into account the fact that the effective separation distance was shorter in the LC device (approximately 2.6 versus 6.2 cm in the microchip-CEC (23) Newman, C. I. D.; Giordano, B. C.; Copper, C. L.; Collins, G. E. Electrophoresis 2008, 29, 803–810.

8290

pump outlet

acetonitrile


50 mM sealed na 50 mM sealed na 50 mM sealed na 50 mM sealed ground

Tris Tris Tris Tris

sample inlet sample in 50 mM open na sample in 50 mM open 2000 V, 100 s sample in 50 mM sealed na sample in 50 mM sealed na

waste Tris Tris Tris Tris

50 mM open na 50 mM open ground 50 mM open na 50 mM open na

Tris Tris Tris Tris

device), the hydrophobicity of the column was enhanced by including phenethyltrimethoxysilane in the precursor solution. Our previous work noted that increasing the alkyl functionality reduced the electroosmotic flow through the stationary phase, though not to the point where separations were impractically long.21 Using an EOF pump as the motive force through a highly functionalized stationary phase may result in more efficient separations and open the door to sol-gel or packed columns that do not support EOF for electrochromatographic separations due to masking of silanol groups on the stationary phase surface. A critical consideration of EOF pump performance in the present application was whether the pump could overcome significant back pressure in order to drive a sufficient flow through the downstream load, a LC column composed of organically modified sol-gel, in this case. The EOF pump performance was characterized by measuring the flow rate of a 20 mM Tris, 60% acetonitrile solution as a function of voltage applied between the pump inlet and pump outlet in the absence and presence of the downstream LC column. The technique used to measure the flow rate has been described previously.19 The flow rate measurements in the absence and presence of a downstream LC column resulted in a linear relationship between flow rate and applied voltage (Without load: flow rates from 20 to 70 nL/min for 1.5-4.5 kV applied, giving a slope of 16.91 nL · min-1 · kV-1 and R2 ) 0.988. With load: flow rates from 15 to 28 nL/min for 2.0-4.5 kV applied, giving a slope of 5.30 nL · min-1 · kV-1 and R2 ) 0.999), indicating no adverse effects due to Joule heating.19 Although Joule heating effects were not expected because of the Tris/acetonitrile solution’s low conductivity, this was confirmed by measuring current as a function of voltage and observing a linear profile. Comparing the flow rates at each applied voltage with and without the downstream LC column, it is noted that the EOF pump was able to drive a flow through the LC column with a flow rate range from 41 to 55% of the load-free flow rate, between an applied voltage from 2.0 to 4.5 kV. Chromatographic Performance. Injection Scheme. Prior to recording chromatograms for the first time, the performance of the electrokinetic pump and the downstream separation process was visually observed by placing the microchip under a fluorescent microscope (Olympus model CKX41) and using the fluorescent dye, Rhodamine B. These efforts helped determine the injection and separation scheme described in the Experimental Section and presented in Table 2. This scheme relies on the analytes being in a purely aqueous 50 mM Tris sample matrix that is injected from

Figure 2. (A) Separation of 1,3,5-TNB and NB at 2.5 and 4.5 kV. Sample was injected electrokinetically from the sample inlet to the waste vial for 100 s at a concentration of 2 µg/mL in 50 mM Tris for each analyte. The pump inlet contained 20 mM Tris, 60% acetonitrile. (B) Relationship between plate height and applied voltage for NB. The pump inlet contained 20 mM Tris, 60% acetonitrile. (n ) 3).

the sample sidearm reservoir directly onto the head of the sol-gel column. The explosives and their degradation products have a high affinity for the stationary phase and, consequently, will extract onto the head of the column. Furthermore, when the sample reservoir was subsequently sealed in preparation for the application of high voltage to the EOF pump, there was no resultant disruption of the sample plug on the head of the sol-gel column. Although sensitivity was not considered a significant performance criterion in this proof-of-concept application, a benefit of the injection scheme used is that relatively large injection periods can be considered for low analyte concentrations in order to load the column head for subsequent elution, as demonstrated by our previous efforts using sol-gels for capillary and microchip electrochromatography.20,21 Effect of Applied Voltage. Once the injection scheme was validated, the performance of the device was determined as a function of the voltage applied to the electroosmotic pump. A mixture of two nitroaromatic explosive degradation products, 1,3,5TNB and NB, were injected and separated using an elution solution of 20 mM Tris, containing 60% acetonitrile. The separation voltage was increased from 2.0 to 4.5 kV in 0.5-kV increments. The resultant separations of 1,3,5-TNB and NB at 2.5 and 4.5 kV are shown in Figure 2A. The 1,3,5-TNB is the first to elute, followed quickly by the NB. As expected, increasing the voltage applied to the EOF pump decreases the migration time of the analytes. The baseline shift observed in both chromatograms is due to the shift in acetonitrile concentration from 0% in the sample matrix (50 mM Tris) to 60% in the elution solution in the pump inlet reservoir (20 mM Tris 60% ACN); this shift can be used to monitor the linear flow velocity of the elution solution through the separation column and will be discussed in more detail later in the paper. An important characteristic of liquid chromatographic separations is the relationship between the height equivalent to a theoretical plate (HETP) and the flow rate of the elution solution. Figure 2B shows the relationship between the HETP of the longer retained NB peak and the applied voltage at the pump. A minimum HETP of 10.4 µm was observed at 3.5 kV, which corresponds to

Figure 3. (A) Two-dimensional plot of the separation of six explosives and degradation products with 4.0 kV applied to the EOF pump. Sample was injected electrokinetically from the sample inlet to the waste vial for 100 s at a concentration of 1 µg/mL in 50 mM Tris for each analyte. The pump inlet contained 20 mM Tris, 60% acetonitrile. (B) Select wavelengths from the 2-D plot demonstrating the usefulness of spectral resolution.

∼99 000 plates/m on a 2.6-cm separation column or 2575 plates (comparing favorably to the results achieved by Fuentes and Woolley, 3350 plates).17 This voltage also provided the shortest HETP for 1,3,5-TNB of 13.4 µm. Resolution between the two peaks ranged from 1.92 to 2.62, with the greatest resolution observed at 3.0 kV. As with the HETP relationship, the resolution and error associated with that measurement were greater at the higher voltages. While an optimum separation with respect to resolution and HETP occurs in the range of 3.0-3.5 kV, further experiments were performed at 4.0 kV to afford a more rapid separation. While resolution is somewhat compromised at this potential, the use of the CCD spectrometer allows for simultaneous temporal and spectral resolution. Figure 3A shows a 2D chromatogram of six nitroaromatic and nitramine explosives and their degradation products using the microfluidic LC device with an elution solution of 20 mM Tris, 60% acetonitrile. While some components are not baseline resolved, analysis of select wavelengths verifies the presence of these analytes (Figure 3B). Effect of Acetonitrile Concentration. An important aspect of a microfluidic device that relies on an EOF pump for the generation of fluid flow is the realization that conditions enabling optimal pump performance (flow rate) and separation efficiency (resolution and time) are not necessarily the same. It is well-known that organic modifiers, in this instance acetonitrile, will strongly affect EOF velocity.24,25 Figure 4A illustrates the effect of acetonitrile concentration [ACN] on electroosmotic flow migration time (time necessary to travel a fixed distance of 9 cm) in a 50-µm-i.d. capillary. As the acetonitrile concentration increases, the migration time for EOF increases, corresponding to slower linear and volume flow rates. When a load is placed on the EOF pump, in this case the sol-gel stationary phase, the migration time of the elution solution behaves differently. Figure 4B shows the corresponding migration times for various analytes in 40-70% acetonitrile through the microchip’s sol-gel stationary phase. These concentrations were (24) Schwer, C.; Kenndler, E. Anal. Chem. 1991, 63, 1801–1807. (25) Wright, P. B.; Lister, A. S.; Dorsey, J. G. Anal. Chem. 1997, 69, 3251– 3259.


8291

concentrations. Clearly, while the flow rate of the elution solution is greatest at the lower acetonitrile concentration, the analytes’ affinity for the stationary phase is greater, resulting in longer retention times. The observed percent covariance in migration time was generally acceptable; for acetonitrile concentrations between 40 and 60%, the variance was less than 5%; however, for the 70% acetonitrile concentration, the variance increased to 15%. This larger error may be due to a combination of the fast flow through the stationary phase and the rapid partitioning into the elution solution.

Figure 4. (A) Relationship between migration time and acetonitrile concentration in an open-tube fused-silica capillary. All solutions contained 20 mM Tris. Length to detector is 9 cm, and total length is 60 cm, 15 kV applied. (n ) 3) B) The relationship between migration time and acetonitrile for the solvent front, RDX, 2-Am-4,6-DNT, NB, and 4-NT in the LC microfluidic device. All solutions contained 20 mM Tris and 4.0 kV was applied to the EOF pump (n ) 3).

examined due to their effectiveness with regard to the separation of explosives. Acetonitrile concentrations less than 30% did not allow for the elution/separation of many of the more retained analytes, while concentrations greater than 70% significantly reduced resolving power to the point that even the use of the CCD spectrometer did not allow for peak identification. From 40 to 50% acetonitrile, the migration time for the solvent increases as the acetonitrile concentration increases, similar to the effect observed in the capillary. As the acetonitrile concentration continues to increase from 50 to 70%, however, the migration time decreases. This behavior is likely attributed to a combination of (1) differences in the wetability of the stationary phase, due to its methyl and phenethyl functionality, and (2) the change in viscosity of the solution with increasing [ACN]. An important caveat of using an EOF pump for on-chip liquid chromatography is that separation optimization must take into account both improvements in pump efficiency and the resultant interaction of the pump fluid with the chemical nature of the stationary phase. A closer examination of Figure 4B indicates that the least retained peak, RDX, has its migration time mirror that of the solvent front. The more retained peaks, including 2-Am-4,6-DNT, NB, and NT, have longer migration times at lower acetonitrile

8292


CONCLUSIONS The integration on a single planar microchip of an EOF pump and a stationary phase within the field-free region resulted in the separation of six nitroaromatic and nitramine explosives/degradation products via liquid chromatography with an HETP as low as 10 µm. The design required a number of fluid dynamic considerations in order to successfully couple the hydraulic performance of the EOF pump with the flow requirements of a packed stationary phase to achieve the separation. These considerations included reducing flow restrictions and back pressure throughout the microchip. This was achieved by increasing the field-free channel cross-sectional area, using a high-porosity sol-gel stationary phase, shortening the weir length, using a packed bed consisting of two bead sizes to allow a deeper weir, and gradually expanding the area between the pressure generating packed channel and the weir. This proof-of-concept design offers a novel integration of an EOF pump, sol-gel stationary phase, and a detection area packaged on a portable planar microchip and offers one example from which further developments of HPLC-based microchips can proceed. ACKNOWLEDGMENT The authors thank the Office of Naval Research for funding support of this effort through the Naval Research Laboratory (NRL). J.F.B. was a NRL Postdoctoral Fellow and sponsored by the American Society for Engineering Education during the course of this work. J.F.B. and B.C.G. contributed equally to this work. Received for review July 17, 2008. Accepted August 26, 2008. AC801497R

Electroosmotic Flow-Based Pump for Liquid Chromatography on a

Recommend Documents