Anal. Chem. 2001, 73, 1974-1978
In-Line Valve Injection for Capillary Electrophoresis Lisa M. Ponton† and Christine E. Evans*,‡
Department of Chemistry, Iowa State University, Ames, Iowa 50011, and Chemical Engineering Department, University of Michigan, Ann Arbor, Michigan 48109-2136
Direct in-line injection is successfully demonstrated for capillary electrophoresis using a commercially available injection valve designed for liquid chromatographic applications. The internal, fluid-contacting materials in this valve injector are composed of ceramics and PEEK (polyetheretherketone). In studies up to 20 kV, this materials design provides a sufficient dielectric interface to insulate the high-voltage buffer from the metal valve body. Partial-loop injections from 6 to >60 nL are shown to be highly reproducible and generally consistent with direct electrokinetic injections under the same experimental conditions. The small extracolumn variance contributed by the valve injection system is symmetrical, and the measured theoretical plates for 75-µm- and 100-µmi.d. separation capillaries are 1.6 × 105 and 2.5 × 105, respectively. As a result, the separation performance is quite good, demonstrating the viability of in-line valve injection for capillary electrophoresis. This development in capillary electrophoretic instrumentation has important implications for the advancement of electrophoretic applications as well as for the design of completely integrated analysis systems. Well-defined and reproducible injection of nanoliter sample volumes remains an enduring challenge in capillary electrophoresis. Most commonly, sample introduction is accomplished using electrokinetic or hydrodynamic means.1 In electrokinetic (EK) injection, analytes are loaded onto the separation capillary using an applied electric field. In hydrodynamic (HD) injection, a small pressure difference introduces the sample volume onto the separation capillary. Although successful, neither technique is as versatile as liquid chromatographic injection schemes in which the sample is introduced directly into the flow stream with minimal interruption of fluid flow. Several advances in injection and coupling techniques have been recently reviewed,2,3 including microloop injection4 and flow-injection (FI) analysis sample * To whom correspondence should be addressed. † Iowa State University. ‡ University of Michigan. (1) Moring, S. E. In Capillary Electrophoresis: Theory and Practice; Grossman, P. D., Colburn, J. C., Eds.; Academic Press: San Diego, CA, 1992; Chapter 3 and references therein. (2) Kuldvee, R.; Kaljurand, M. Crit. Rev. Anal. Chem. 1999, 29, 29-68. (3) Veraart, J. R.; Lingeman, H.; Brinkman, U. A. Th. J. Chromatogr. A 1999, 856, 483-514. (4) Dasgupta, P. K.; Surowiec, K. Anal. Chem. 1996, 68, 1164-1168.
1974 Analytical Chemistry, Vol. 73, No. 9, May 1, 2001
introduction, as discussed in refs 5-8 and references therein. Both of these methods provide the small injection volumes that are necessary and with precision values rivaling EK and HD injection. Of these, FI is perhaps the most similar to a direct in-line injection scheme. In FI-CE, samples are introduced into the flow injection system using a larger volume valve, and the process proceeds without any interruption of the applied voltage. As a result, FI-CE can be integrated into a wide range of analysis configurations. In contrast with direct-injection methods, sample injection by FI uses pump-driven flow within the FI system, and the inlet to the CE system must be grounded. In addition, by the very nature of FI, sample introduction is downstream and necessarily results in some sample dilution. That said, FI and direct methods for sample introduction are clearly complementary, each providing distinct advantages. Unfortunately, reports of direct in-line sample introduction have been very limited.9,10 Materials requirements have hindered the use of liquid chromatographic injectors because of bubble formation within the metal valve structure, which interrupt the electrical current and fluid flow. Nonetheless, sample injection using a rotary-valve injector was introduced as early as 1987.9 Fabricated from tetrafluoroethylene resin and ceramics, this valve provided injection reproducibilities of ∼2%, which was consistent with EK and HD methods. Although this early design suffered from relatively large sample volumes (∼350 nL), it is unclear why this injection method was not further exploited. In this manuscript, we describe the successful implementation of direct in-line sample introduction for capillary electrophoresis using an injection valve that is presently commercially available for liquid chromatography. With the exception of safety features noted later in the paper, no further modification was made to the commercial valve. As a result, the valve design was not optimized for CE and the data represent the commercial valve performance. EXPERIMENTAL METHODS Chemicals. Mesityl oxide and the positional isomers of o-, m-, and p-nitrophenol were obtained from Aldrich (Milwaukee, WI) and used without further purification. Sodium hydroxide and sodium borate decahydrate (Na2B4O7‚10H2O), used for cleaning and buffering respectively, were purchased from Sigma (St. Louis, (5) Fang, Z.-L.; Lui, Z.-S.; Shen, Q. Anal. Chim. Acta 1997, 346, 135-143. (6) Fang, Z. L.; Chen, H. W.; Fang, Q.; Pu, Q. S. Anal. Sci. 2000, 16, 197-203. (7) Kuban, P.; Karlberg, B. Anal. Chim. Acta 2000, 404, 19-28. (8) Melanson, J. E.; Lucy, C. A. J. Chromatogr. A 2000, 884, 311-316. (9) Tsuda, T.; Mizuno, T.; Akiyama, J. Anal. Chem. 1987, 59, 799-800. (10) Evans, C. E. Anal. Chem. 1997, 69, 2952-2954. 10.1021/ac001145r CCC: $20.00
© 2001 American Chemical Society Published on Web 03/28/2001
Table 1. Capillary Electrophoretic Conditions for Separation Capillaries of 75 and 100 µm I.d. 75-µm i.d.
Figure 1. Instrumental schematic representation of injection valve ports and cross-sectional view of the injection valve. While the valve is in the load position, the buffer flows onto the separation capillary, bypassing the sample loop. In the inject position, buffer flows through the sample loop, sweeping sample onto the separation capillary by electroosmotic flow. The cross-sectional view of the injection valve shows two capillaries in contact with a moveable channel by means of two connectors (not drawn to scale). Each of the three rotor channels has dimensions of width, 0.125 mm; length, 2.1 mm; depth, 0.137 mm; and a total volume of 36 nL. Each of the six fixed connectors has dimensions of length, 0.25 mm and diameter, 0.22 mm, and a volume of 9.5 nL.11
MO). All solutions were made using distilled, deionized water (Millipore; model Milli-Q UV Plus; Bedford, MA; >18 MΩ). Standard solutions were prepared in the running buffer at ∼0.1 mM for each solute. Instrumental Design. The injection valve used in these experiments is a 6-port liquid chromatography micro-injector valve (model M-435; UpChurch Sci.; Oak Harbor, WA). The internal materials are manufactured from ceramics or PEEK; as a result, the metal injector body is electrically insulated from the highvoltage buffer. This allows the high-voltage power supply to be kept on throughout the injection and separation process. As illustrated in Figure 1, each channel within the rotor has a connector region at each end, which brings the channel into contact with the capillaries. The three moveable channels within the rotor assembly are 36 nL each, with the six fixed connectors of 9.5 nL each. In the load position (Figure 1), two channels and the sample loop are filled with sample from a syringe. With the electric field on, partial-loop injection is accomplished by manually switching to the inject position for a set amount of time, then returning to the load position. Although the voltage is continuously supplied to the separation capillary, valve switching partially interrupts the applied voltage for the very short interval when the rotor is between ports. As with all electrophoresis experiments, safety is of paramount importance. In this study, all of the high-voltage connections were contained within a Plexiglas enclosure. Throughout all of the studies with the valve, measurements under high-voltage conditions indicated no voltage leakage to the valve body (the valve body was not grounded). Applied voltage conditions for capillary electrophoresis separations of up to 20 kV were assessed, which resulted in a maximum voltage at the valve of 5-6 kV. For the studies shown here, voltage values at the valve were more typically ∼3 kV. Safety precautions were instituted to further isolate and protect the operator in case of failure. The valve was mounted on
100-µm i.d.
EK injections tot length, cm eff length, cma appl volt, kV elec field, V/cm meas curr, µA
93.4 61.8 11.275 120.7 15.8
95.5 62.1 11.531 120.7 29.5 µA
valve injections tot length, cm eff length, cm sample loop, cm buffer cap, cm sep cap, cm appl volt, kV elec field, V/cm meas curr, µA
120.0 62.0 8.5 (0.38 µL) 26.6 93.4 14.483 120.7 15.6
120.3 62.3 8.5 (0.67 µL) 26.8 93.5 14.524 120.7 32.5
a
Effective length from injection to detection.
the wall of the insulating Plexiglas enclosure, and all voltage and liquid connections were located inside the box. Another Plexiglas box was built around the small portion of the valve that extended outside the larger enclosure. For safe manual operation, the metal injector handle supplied with the valve was replaced with a sixinch Teflon extension. As with the valve body, measurements showed no voltage leakage to the liquid sample introduction or waste under high-voltage conditions. Nonetheless, all liquid connections to the valve for filling and cleaning the internal loops used Teflon components and fused-silica capillaries to ensure insulation from any residual voltage. Using these precautions ensured that operation of the experiment proceeded safely and without difficulty. As shown in Figure 1, five lengths of capillary were used to connect fluid flow: separation capillary, sample loop, ground buffer, sample waste, and sample inlet. All capillary lengths and experimental conditions for these studies are noted in Table 1. As shown, experimental conditions remained identical for valve and electrokinetic injections to the extent possible. For the valve injections, the separation capillary extended from the injector to the negative high-voltage buffer vial. An additional capillary length was used to connect the grounded inlet buffer vial to the injector. In both the load and inject positions, fluid flow proceeded between buffer vials and through the injector onto the separation capillary. Sample loading was accomplished from a syringe coupled to a capillary, through to the injector, a capillary sample loop, and out a capillary to the waste vial. Injector performance was assessed for separation capillaries of 75- and 100-µm i.d., and the applied voltage was adjusted to maintain the electric field constant (Table 1). The applied electric field was controlled throughout these studies using a negative-voltage power supply (model MJ30N0400; Glassman; Whitehouse Station, NJ). UV detection (model UVIS-205; Linear; Chicago, IL) at 225 nm was performed directly on-line with computer data acquisition (LabView, version 3.1.1). Electrophoretic Conditions. All electrophoretic conditions are summarized in Table 1. Capillaries were prepared each day by rinsing all capillaries and the injector flow paths successively with 0.1 N NaOH, high-purity water, and buffer solution. A 25 Analytical Chemistry, Vol. 73, No. 9, May 1, 2001
1975
Table 2. Measured Reproducibility and Plate Height for Electrokinetic and Valve Injectionsa,b 75 µm i.d. inj time, s
100 µm i.d.
inj vol, nL
height, %
area, %
H, µm
inj vol, nL
height, %
area, %
H, µm
2 4 8 16
5.2 10.4 20.4 42.3
10.1 2.1 1.7 2.3
10.9 3.5 2.6 3.8
3.4 (0.3) 4.0 (0.3) 7.8 (0.4) 28 (1)
10.1 20.1 39.9 NA
8.7 1.9 1.6 NA
10.2 1.8 2.6 NA
3.65 (0.03) 4.5 (0.6) 8.2 (0.5) NA
valve 2 5 10
6.3 19.4 42.2
7.7 10.3 0.9
5.8 14.3 1.5
7.0 (0.5) 8.3 (0.7) 11.90 (0.04)
9.7 33.6 66.6
2.9 3.9 2.8
1.6 4.2 4.9
EK
4.9 (0.4) 6.0 (0.4) 10.6 (0.1)
a Measured as relative standard deviations in the peak height and area. b All values are based on mesityl oxide injections (n g 5). Electrokinetic (EK) injection volumes were determined by electroosmotic flowrate and injection time; valve injection volumes were determined by the peak area calibration curve of EK injection volumes. Height and area are shown as relative standard deviations. Average plate height is given with standard deviation in parentheses.
mM borate solution (pH ) 9.5) was used for all measurements. Ambient temperature was maintained at 20 ( 1 °C throughout these studies and was not further controlled. RESULTS AND DISCUSSION In capillary electrophoresis separation techniques, sample introduction is accomplished nearly exclusively off-line. The primary injection methods in present use, electrokinetic and hydrodynamic techniques, require that the separation capillary to be physically moved to a sample vial. We introduce here the implementation of a commercially available valve for in-line sample injection in capillary electrophoresis. In this injection method, the applied electric field need not be turned on and off during the injection process, and the sample is introduced directly into the separation capillary using a rotary valve injector. The use of ceramics or PEEK for all of the internal valve materials renders the flow channel insulated from the stainless steel valve body. Although the present valve is designed for liquid chromatographic applications, an intermittent or partial-loop injection process can be utilized to inject the small sample volumes that are required for capillary electrophoresis methods. Injection System. Similarly to liquid chromatographic applications, the injection valve is positioned directly in-line. However, instead of a pump driving the fluid flow, external voltage is constantly applied to the separation capillary. In the load position (Figure 1), a syringe is used to fill the sample loop with the sample of interest. Upon switching to the inject position, the sample loop is moved in-line, and the sample is electroosmotically driven onto the separation capillary for a given time interval. The valve is then returned to the load position, and the electrophoretic separation proceeds without any physical movement of the separation capillary. In contrast with more conventional injection methods, a systematic error in the injection volume delivered by the valve is observed when the determination is made using the neutral marker migration. This likely arises from differences between the flowpath during injection and that during separation (i.e., E field and channel diameters). As a result, injection volumes for the valve are determined based on peak-area calibration curves that are derived from direct electrokinetic injections. Injection Reproducibility. Evaluation of this injection valve for CE centers on two primary considerations: injection reproduc1976 Analytical Chemistry, Vol. 73, No. 9, May 1, 2001
ibility and separation efficiency. As shown in Table 2, reproducibility values for the valve injections generally compare favorably with electrokinetic injections (EK) under the same conditions (Table 1). As expected, there is a general tradeoff between injection volume and reproducibility, with manual timing as the limiting factor for the fastest times. Nonetheless, with the possible exception of the 5-s injections for the 75-µm-i.d. capillary, injections are generally reproducible and consistent with those achieved by direct electrokinetic injection. Separation Efficiency. The other primary factor of importance in assessing the performance of a new injection system is the separation efficiency. Because separation efficiencies are very high for capillary electrophoretic separations, all extracolumn sources of variance must be kept to a minimum. This includes the contribution of the injection process to the total efficiency. As shown in Table 2, expected trends of measured plate height with injection volume are observed. Measured plate height values are systematically greater for valve injections than electrokinetic (EK) injections (Tables 2 and 3). Consistent with expectations, better agreement is observed for the 100-µm-i.d. separation capillary, in which extra-column constraints are somewhat relaxed. The standard deviation of the measured plate height values is also consistent with the good reproducibility in peak height and area shown in Table 2. Moreover, the plate height increase that is expected with elution time is observed for the elution order of m-nitrophenol, p-nitrophenol, and then o-nitrophenol. Overall, the valve injection process appears to be contributing additional extracolumn variance. Several models are examined here to evaluate the potential origin of the extracolumn contribution.12 Potential limiting case contributions include (1) diffusion throughout the entire valve, (2) laminar flow within the valve, or (3) new spreading within the valve. Exponential mixing or diffusion chamber contributions can be eliminated, because the exponential contribution from these sources would cause peak asymmetry, which is not observed. Each of the remaining contributions is evaluated below, and the predictions are summarized in Table 3. All plate height predictions are for the neutral marker and include the injection volume (11) Specific valve dimensions obtained from personal communication with John W. Batts, IV, of UpChurch Scientific (Oak Harbor, WA). (12) Sternberg, J. C. Adv. Chromatogr. 1966, 2, 205-270.
Table 3. Measured and Predicted Plate Height Values (µm) for Electrokinetic (EK) and Valve Injections of 10 nL. measureda EK
predictedb
valve
cap i.d.
MOc
MOc
o-NPc
m-NPc
p-NPc
diffusion
laminar
spreading
75 µm 100 µm
4.0 (0.20) 3.65 (0.03)
7.3 (0.5) 4.8 (0.3)
10.6 (0.8) 7.3 (0.2)
8.9 (0.4) 6.4 (0.2)
9.8 (0.4) 7.0 (0.6)
3.3 2.7
3.4 2.8
8.3 4.3
a Measurement conditions given in Table 1 and standard deviations are given in parentheses (n g 5). b Predicted values for mesityl oxide based on the injection volume contribution and a measured diffusion coefficient of 7.5 × 10-6 cm2/s. Additional contributions described in text. c MO: mesityl oxide; o-NP, m-NP, p-NP: the isomers of nitrophenol.
contribution (assuming an idealized rectangular profile of 10 nL) as well as the diffusion contribution through the separation capillary to the point of detection.12 The volume and time constant for the detector have been minimized and are not significant contributors. Variance contributions are assumed to be independent and additive for all calculations.12 Case 1: Diffusion within the Valve. Assuming that the only other source of variance is diffusion within the valve pathway, plate height predictions are based on the Einstein equation (σ2L ) 2Dt), where the length variance (σL2) is simply predicted from the diffusion coefficient (D) and the residence time (t). A measured diffusion coefficient of 7.5 × 10-6 cm2/s is used for all predictions. As shown in Table 3, the values for plate height (H ) σL2/L) predicted for this diffusion-only mechanism are in reasonable agreement with the measured values for direct electrokinetic injection. Thus, the contribution of diffusional variance arising within the valve is not significant, and this value represents the injection volume and separation capillary diffusion contributions. This diffusion-only origin significantly underestimates the measured plate height from the valve measurements. Case 2: Laminar Flow within the Valve. Assuming that laminar flow is present within the valve flow pathway, this contribution is described by σL2 ) L[(2D/u) + (r2u/24D)], where L is the length, r is the radius of the channel, and u is the linear velocity within the flow path. As shown in Table 3, laminar flow within the valve contributes little more than the diffusion-only mechanism and, as a result, still underestimates the measured plate height. Case 3: New Spreading within the Valve. Assuming that socalled new spreading within the valve contributes to the plate height, the variance contribution can be estimated from the limiting case in which the length variance remains constant upon entering the larger diameter connector (d2).12 In this case, the time variance in (σt2in) and out (σt2out) of the larger diameter chamber is given by σt2out/σt2in ) (d2/d1)4. This broadening mechanism occurs when the momentum of flow from a smaller diameter tube (the valve channel) acts to maintain the length variance as the band enters the larger diameter tube (the valve connector). This mechanism requires that lateral mass transfer within the larger tube is immediate and, therefore, is only a limiting case. Of interest here, this broadening could arise within the valve connector regions before and after the valve channel (Figure 1). For the purposes of calculation, the valve channel is approximated as a cylindrical tube. As shown in Table 3, plate height predictions for this mechanism are in closest agreement with the measured plate height values for the valve injection. As
a result, this symmetrical, new spreading mechanism may be a primary contributor to the observed extracolumn variance for the valve injector. Future Improvements. This injection valve provides reproducible and reliable sample introduction for CE as it is presently commercially available. Nonetheless, it has not been optimized for CE, and several improvements can be envisioned. In the present design, the volume of the external sample loop necessitates partial-loop injections for CE. By modifying this design to an internal loop injector, fewer flow transitions will be required, and the timing variability inherent in partial-loop injections can be minimized. Further improvement would be realized by incorporating computer-controlled actuation instead of manual actuation. This would increase the reproducibility of any partial injection timing, if necessary, and better control the partial interruption of the electric field during valve switching, leading to improvements in the overall injection reproducibility. Finally, any new spreading contribution to extracolumn variance could be minimized by closely matching channel diameters within the injection region. Implications. In this paper, we successfully demonstrate the use of a commercially available LC injection valve for in-line injection in capillary electrophoresis. Reproducible sample introduction directly into the high-voltage region on the separation capillary is realized, with only small additional band broadening. This direct in-line sample introduction scheme has several compelling advantages over present methods, including ease of operation and automation, minimal voltage change during sample introduction, and increased versatility relative to existing methods. Moreover, this injection technique can provide the rapid sample repetition rates necessary for two-dimensional separations and for the reaction-monitoring that is of importance for biotechnology applications. In addition, by introducing the sample directly inline, separations can now be optimized in both directions along the capillary. The distinct advantages of such in-line accessibility can be envisaged for the wide range of electrophoretic separation modes, including isotachophoresis. Because the sample can be introduced at any point along the separation capillary, either end of the separation capillary can be placed at high voltage, thus increasing the versatility for detection schemes that require the outlet to be grounded (e.g., electrochemical, mass spectrometry, and conductivity). Clearly, these advantages are not limited to sample introduction, and may also be exploited for applications requiring in-line switching and shunting, as for on-line derivatization. Although demonstrated here using syringe injection, this in-line scheme may be directly incorporated into an integrated Analytical Chemistry, Vol. 73, No. 9, May 1, 2001
1977
systems design using a complete range of flow schemes, including electrically driven flow. This capability also extends to the design of closed-system electrophoretic separations, where it is important to isolate the separation process from the outside environment. This capability becomes crucial in handling extremely hazardous biomaterials or air-sensitive samples. Taken together, these examples illustrate the potential of this injection scheme to significantly advance electrophoretic separations and integrated systems design.
1978
Analytical Chemistry, Vol. 73, No. 9, May 1, 2001
ACKNOWLEDGMENT The authors kindly acknowledge the National Science Foundation (Grant CHE-9707513) for support of this work and John W. Batts, IV of UpChurch Scientific for technical assistance with valve specifications. Received for review September 25, 2000. Accepted February 21, 2001. AC001145R