Anal. Chem. 1998, 70, 2248-2253
Minimization of Sample Discrimination Introduced by On-Column Fracture/Electrokinetic Injection in Capillary Electrophoresis Honging Wei,† Koon Chye Ang,‡ and Sam F. Y. Li*,†,§
Department of Chemistry, National University of Singapore, CE Resources Pte. Ltd., NUS Innovation Center, and Institute of Materials Research and Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260, Republic of Singapore
Using on-column fracture/electrokinetic injection for sample introduction in capillary electrophoresis (CE) is thought to be a method of no-discrimination electrokinetic injection. However, in this study we found that significant discrimination was observed when injecting samples dissolved in deionized water with the above method. In addition, the discrimination is reverse to that in conventional electrokinetic injection, that is, the less mobile species are injected in larger quantities than the more mobile components. The reason for these phenomena and approaches to reduce the discrimination were studied. Equivalent circuits were established and used to analyze the discrimination under different conditions through computer simulation. The experimental results showed good agreement with the results of the computer simulations. Finally, an on-column fracture/electrokinetic injection method using ramped injection voltage was proposed and shown to be suitable for sample solutions with different conductivity. This method does not cause significant error for practical quantification, even without correcting for the discrimination, because its discrimination is very small. Capillary electrophoresis (CE) is noted for its high resolution and small sample volume.1 To preserve the high efficiency capability of CE, it is very important to introduce a small volume (∼10 nL) of sample into the capillary precisely and reproducibly.2-4 Currently, the most commonly used methods are on-column hydrodynamic and electrokinetic injection techniques.5 In both of these types of injection methods, one end of the capillary is used as the sample injector directly. A major limitation of the hydrodynamic methods, including pressure, vacuum and gravity injection, is that they are not suitable †
Department of Chemistry, National University of Singapore. CE Resources PTe. Ltd. § Institute of Materials Research and Engineering, National University of Singapore. (1) Jorgenson, J. W.; Lukacs, K. D. Anal. Chem. 1981, 53, 1298. (2) Huang, X.; Coleman, W.; Zare, R. A. J. Chromatogr. 1989, 480, 95. (3) Lauer, H.; McManigill, D. Trends Anal. Chem. 1986, 5, 11. (4) Grushka, E.; McCormick, R. J. Chromatogr. 1989, 471, 421. (5) Li, S. F. Y. Capillary Electrophoresis-Principles, Practice and Applications; Journal of Chromatography Library, Vol. 52; Elsevier: Amsterdam, 1992; pp 31-53. ‡
2248 Analytical Chemistry, Vol. 70, No. 11, June 1, 1998
for the injection of highly viscous samples or for capillary gel electrophoresis and capillary electrochromatography, due to the fact that hydrodynamic flow would be hampered or suppressed.5 Electrokinetic injection can be utilized in these cases. However, conventional electrokinetic injection is biased in how it introduces differently charged analytes.5-6 Two types of discrimination may occur. One occurs as a result of differences in mobilities of the species in the sample solution. The more mobile components are injected in larger quantities than the less mobile species. Another type of discrimination is related to the differences in the conductivity between the sample solution and the operating buffer. Huang et al. 6 showed for cations (K+ and Li+) that electrokinetic injection introduces a linear bias in which more ions are injected for solutions having higher ohmic resistance. Although there are many ways to correct for these errors,6-7 they make practical quantification complicated, if not inaccurate. Developing methods for no-discrimination electrokinetic injection would be very useful. In the development of methods toward no-discrimination electrokinetic injection, use of an on-column fracture for sample introduction 8 appeared quite promising. In this method, an oncolumn fracture was made on the capillary as that for electrochemical detection.9 The on-column fracture allows for ions and therefore, current to pass, but does not allow a significant quantity of buffer solution to pass through. During injection, a potential is applied between the fracture and the outlet of the capillary. Electroosmotic flow (EOF) pulls sample ions into the capillary through the inlet of the capillary. The amount of sample introduced is proportional to the EOF. It is assumed that no electromigration of analytes occurs and introduction of sample is no longer biased by using only the EOF as pump.8 In this sense, this method may be called electrically driven hydrodynamic injection. However, we find that discrimination still exists when injecting samples dissolved in deionized water with the above device. In addition, the discrimination is reverse to that in conventional electrokinetic injection, that is, the less mobile species are injected in larger quantities than the more mobile components. In this study, the reason for these phenomena and approaches to reduce (6) Huang, X.; Gordon, M.; Zare, R. A. Anal. Chem. 1988, 60, 375. (7) Qi, S.; Huang, A.; Sun, Y. Anal. Chem. 1996, 68, 1342-1346. (8) Linhares, M. C.; Kissinger, P. T. Anal. Chem. 1991, 63, 2076-2078. (9) Wallingford, R. A.; Ewing, A. G. Anal. Chem. 1988, 60, 258. S0003-2700(97)00951-7 CCC: $15.00
© 1998 American Chemical Society Published on Web 05/01/1998
Figure 1. The assembly of the fracture and buffer reservoir.
the discrimination were studied. Equivalent circuits were established and used to analyze the discrimination under different conditions through computer simulation. The experimental results showed good agreement with the results of the computer simulations. EXPERIMENTAL SECTION Apparatus. The CZE system was based on an in-house design. A personal computer combined with a high-voltage power supplier (Spellman High Voltage Electronics Corp., NY) was used to control the voltage for electrophoresis. On-column detection was performed at 210 nm with a model 200 UV detector (Linear Instruments, NV). Data acquisition was achieved also by the PC. Untreated fused-silica capillaries (50 µm i.d. × 360 µm o.d., 50 cm in length with 40 cm in effective length) were purchased from Polymicro Technologies (Phoenix, AZ) and used for separation. Software used for simulation was Simulink under Matlab (version 4.2.c.1, The Mathworks, Inc.) Construction of On-Column Fracture. The procedure is nearly the same as that previously described.8 Briefly, the polyimide coating about 6 cm from the inlet of the separation capillary was burned off first. Then epoxy was applied on both sides of the exposed part of the capillary to glue it on a 0.5 cm × 1.0 cm microscope slide. A small scratch was made on the top of the exposed part after the epoxy was totally dried. Finally, an on-column fracture was made through pushing up gently from the bottom, directly under the scratch, with a pointed stylet. The fracture assembly was then placed in a 5-mL disposable syringe tube shortened to the required length. The short end of the capillary was pushed through a rubber septum that plugged the hole in the bottom of the tube. Then the tube was filled with running buffer. A 24-gauge platinum wire electrode was inserted into the tube. This apparatus is shown in Figure 1. Sample Introduction Procedure. In our experiments, we used either two platinum electrodes (one in the tube, another one in the outlet buffer reservoir) or three platinum electrodes (one in the tube, one in the sample vial, and another one in the outlet buffer reservoir) to introduce the sample. In the first case, we will call it two-electrode injection. The procedure employed was
similar to that described previously.8 During injections, the power supply was connected to the electrodes in the fracture assembly and the outlet reservoir (switch 2 was on while switch 1 was off, see Figure 1). The inlet end of the capillary was placed into the sample vial. Sample injections were timed by a stopwatch. The solution levels in the inlet and outlet reservoirs were maintained at the same height to avoid siphoning. After injection, the sample vial was replaced with a buffer reservoir. In the second case, we will call it three-electrode injection. The procedure is nearly the same as in the first case except that both the electrodes in the tube and the sample vial were connected to the positive output of the power supply (both switches 1 and 2 were on) during injection. Gravity injection, one of hydrodynamic methods, was chosen to compare with the electrokinetic injection method. To perform gravity injection, the inlet of the capillary was inserted into the sample vial and kept at 2 cm height over the outlet of the capillary for a certain time. During run conditions, the power supply was connected to the electrodes in the inlet and the outlet reservoirs (switch 1 was on while switch 2 was off), and 15 kV (300 V/cm) was applied across the entire capillary. Discrimination Determination. A positive ion, benzylamine, a negative ion, benzoic acid, and a neutral marker, dimethyl sulfoxide (DMSO) were chosen as sample components in neutral aqueous solution. After each run, the ratios of integrated area of benzylamine and benzoic acid to DMSO were determined. If discrimination exists, the area ratios obtained by the electrokinetic injection will be different from that obtained by hydrodynamic injection for the same sample solution. For example, the area ratio of benzylamine to DMSO obtained by conventional electrokinetic injection is greater than that obtained by hydrodynamic injection because the mobility of benzylamine is greatest among three sample components. Reagents. Electrophoresis buffer was 80 mM phosphate buffer (pH 7.0). It was prepared by using deionized water generated by a D4700 Nanopure water-purification system (Barnstead/Thermdyne Corp., IA) and filtered through a 0.45 µm nylon membrane before use. Samples in three solutions with different conductivity were prepared by dissolving the three sample components in different solutions. For the sample solution with low conductivity, only deionized water was used to dissolve the sample components. Running buffer was used for the sample solution with medium conductivity, while for the sample solution with high conductivity, running buffer with 0.2 M NaCl was used. Phosphate salts, DMSO, benzylamine, and benzoic acid were bought from Fluka (Buchs, Switzerland) and were used as received. RESULTS AND DISCUSSION Characterization of Discrimination in Sample Introduction by On-Column Fracture/Electrokinetic Injection. As mentioned in the Introduction, two types of discrimination exist in conventional electrokinetic injection. To investigate if an electrokinetic injection method is biased, it is necessary to compare the area ratios of different mobile components introduced by electrokinetic injection from sample solutions with different conductivity to that introduced by hydrodynamic injection. If no difference between the two methods could be found in all cases, we would conclude that no discrimination exists. Analytical Chemistry, Vol. 70, No. 11, June 1, 1998
2249
Figure 2. Electropherograms for benzylamine (1), DMSO (2), and benzoic acid (3) introduced by gravity injection (A) and by twoelectrode injection under different injection voltages (B, 5 kV; C, 10 kV). Running buffer: 80 mM phosphate buffer (pH 7.0). The distance from the inlet of the capillary to the fracture is 6.5 cm. In A, gravity injections were performed at 2 cm height × 60 s. Sample solution was prepared by dissolving sample components in deionized water, and the concentrations were about 100, 720, and 254 ppm for benzylamine, DMSO, and benzoic acid, respectively. Separation voltage, 265 V/cm. The numbers above the peaks are percentages of the integrated area of the corresponding peak. The numbers between peaks 1 and 2 are the area ratios of benzylamine to DMSO. The numbers between peaks 2 and 3 are the area ratios of benzoic acid to DMSO.
We tested the method previously mentioned8 (see Experimental Section, two-electrode electrokinetic injection) with the three different sample solutions first. For the samples dissolved in buffer and in high-conductivity solution, no deviations among the area ratios introduced by two-electrode electrokinetic injection and that introduced by gravity injection were found within experimental errors. However, significant deviations were obtained when introducing sample from the solution with low conductivity, that is, samples in deionized water, as shown in Figure 2. We can see from Figure 2 that the area ratio of benzylamine to DMSO introduced by two-electrode electrokinetic injection was less than that introduced by gravity injection, while the area ratio of benzoic acid to DMSO was significantly larger. These results are the opposite of that expected in conventional electrokinetic injection.10-11 In addition, the lower the injection voltage, the less the deviation. (10) Liu, J. H.; Chung, Y. C. Angew. Makromol. Chem. 1994, 219, 101-115. (11) Dose, E. V.; Guiochon, G. A. Anal. Chem. 1991, 63, 1154-1158.
2250 Analytical Chemistry, Vol. 70, No. 11, June 1, 1998
The increase of discrimination is also verified by increasing retardation of the peak of benzylamine. These phenomena reminded us that an electric field existed at the interface between the sample solution with low conductivity and the solution in the capillary (running buffer) during injection. We believe that it is possible for the unexpected behavior to occur after considering the injection process further as discussed below. Explanation of the Discrimination. Before the injection voltage is applied between the tube and the outlet reservoir, both the electrical potentials of the sample solution and the running buffer are equal to zero. When the injection voltage is applied, the potential of the running buffer at the fracture will rapidly increase to the injection voltage. At the same time, the potential of the sample solution, like a floating end in an electric circuit, will increase also until its potential approaches that of the running buffer adjacent to it. It is well-known in fundamental electronics that an isolated metal sphere possesses a capacitance.12 Since the sample solution and the running buffer consist of ions and can conduct current, the sample solution and the buffer between the inlet and the fracture of the capillary could behave as an isolated metal sphere in this situation and should possess a capacitance. Because of this capacitance, the potential of the sample solution will not rise equally with the increase in the potential of the buffer when the power supply is on (like a charging process of a capacitor12). This means that the sample solution will have a lower potential than the running buffer, and there exists an electric field between the sample solution and the running buffer in the capillary. This electric field can cause positive ions to move away from the inlet of the capillary and negative ions to move into the capillary, and thus the less mobile components are injected in larger amounts than the more mobile components. Time constant (RC, resistance × capacitance) is an important factor for describing a charging process. In this case, the time constant will be the resistance of the liquid (the sample solution and the buffer between the inlet and the fracture of the capillary) times the capacitance. The bigger the time constant (RC), the larger the potential difference between the sample solution and the running buffer.12 Therefore, larger discrimination could be observed. For the sample solution dissolved in deionized water, its resistance is quite high. Hence, its RC is large. We can expect that the discrimination will be larger than those in the cases of the other two sample solutions. This observation is consistent with the sample stacking effect in field-amplified sample injection (FASI) proposed by Chien and Burgi,13,14 where the sample discrimination was found to depend on the γ-ratio of buffer concentrations in the column to that in the original sample solutions. After the charging process, the potential of the sample solution will be equal to that of the running buffer. The discrimination will not exist. At the same time, the interface between the sample solution and the running buffer moves into the capillary when more sample solution is pumped into the capillary by EOF. When the power supply is off at the injection end, a depolarization process will occur. However, this process would have little effects on the total amount of the sample injected into the capillary (12) Oldham, H. B.; Myland, J. C. Fundamentals of Electrochemical Science; Academic Press: San Diego, CA, 1994; p 26. (13) Chien, R. L.; Burgi, D. S. J. Chromatogr. 1991, 559, 141-152. (14) Chien, R. L.; Burgi, D. S. Anal. Chem. 1991, 63, 2042-2047.
Figure 4. (A) The electropherogram for bezylamine (1), DMSO (2), and benzoic acid (3) introduced by three-electrode injection (injection voltage, 10 kV for 4 s). For other conditions, see Figure 2. (B) The equivalent circuit for the three-electrode injection. R2, the resistance of the sample solution and the running buffer between the inlet and the fracture of the capillary; R1, the resistance of the running buffer between the fracture and the outlet of the capillary, Rf, the resistance across the fracture; R3, the resistance of the buffer between the Pt electrode and the fracture in the tube.
Figure 3. (A) The equivalent circuit for the two-electrode injection. R1 is the resistance of the sample solution and the running buffer between the inlet and the fracture of the capillary, C is the capacitance of the above liquid, R2 is the resistance of the running buffer between the fracture and the outlet of the capillary, Rf is the resistance across the fracture, and R3 is the resistance of the buffer between the Pt electrode and the fracture in the tube. (B) Computer simulation results of ∆V under different injection voltages during the charging process of the equivalent circuit in (A) (RC ) 0.1 s). Curve 1, 600 V; curve 2, 300 V; curve 3, 100 V; curve 4, 30 V. (C) Computer simulation results of ∆V under different time constants during the charging process of the equivalent circuit in (A) (injection voltage, 600 V). Curve 1, RC ) 0.1 s; curve 2, RC ) 0.05 s; curve 3, RC ) 0.01 s; curve 4, RC ) 0.005 s.
since the interface between the sample solution and the running buffer is now in the capillary. Therefore, we can only consider the effect of the charging process on the discrimination. An equivalent circuit, shown in Figure 3A, could be used to analyze the charging process. The potential difference, ∆V, is the reason for the discrimination. Integration of ∆Vdt during the charging process can be used to estimate the value of the discrimination. In the following sections, we will focus on how to reduce the discrimination when injecting samples from the sample solution with low conductivity since no significant discrimination could be observed in the cases of the other two sample solutions. Three-Electrode Injection. In our further experiments, we tested the use of the three-electrode electrokinetic injection method to see if the discrimination could be eliminated since no liquid will be isolated. The results are shown in Figure 4A. The bias was reversed and behaved like a conventional electrokinetic injection. The reason may be that there exists an electric field between the inlet of the capillary and the fracture during the injection process. An equivalent circuit for this situation was shown in Figure 4B. Due to resistance of R3 and Rf, there is a Analytical Chemistry, Vol. 70, No. 11, June 1, 1998
2251
Figure 5. (A) The change of the injection voltage with time during the ramped two-electrode injection. (B) Computer simulation results of ∆V under different ramped rates during the charging process of the equivalent circuit in Figure 3 with a ramped injection voltage (maximum injection voltage, 600 V). Ramped rates: curve 1, k ) 600 V/s; curve 2, k ) 200 V/s; curve 3, k ) 100 V/s; curve 4, k ) 66.7 V/s.
potential drop over R2 (the sample solution and the running buffer between the inlet of the capillary and the fracture). Computer Simulations. Based on the equivalent circuits in Figure 3A, we can use a computer to simulate how to reduce the discrimination introduced by on-column fracture/electrokinetic injection. We will only consider the charging process since it is the only process to produce the discrimination during injection. In our experiments using two-electrode injection, we found that lower injection voltage could result in smaller discrimination. Therefore, we will focus on how the injection voltage affects the discrimination. Figure 3B shows the simulated effects of the injection voltage on ∆V using the equivalent circuit in Figure 3A. We can see that the lower the injection voltage, the less the potential difference will be. This means that the discrimination will decrease with decrease in the injection voltage, which is in agreement with previous experimental results. Figure 3C shows the change of ∆V under different time constants for simulating the two-electrode injection. We can see ∫∆Vdt decreases with decrease in the time constant, which means that the discrimination will become smaller with decrease in the time constant. This may explain why significant discrimination was observed only when the sample was 2252 Analytical Chemistry, Vol. 70, No. 11, June 1, 1998
Figure 6. The electropherograms for benzylamine (1), DMSO (2), and benzoic acid (3) introduced by ramped three-electrode injection. (A) Ramped rate, k ) 1 kV/s. (B) k ) 166 V/s. Maximum injection voltage, 5 kV. Other conditions are the same as those in Figure 2.
dissolved in deionized water, since its time constant was the biggest among the three sample solutions. From the above simulation results, we can conclude that reducing injection voltage can reduce the discrimination. However, we could not use a very small injection voltage since it will take a long time to inject enough sample solution (∼10 nL) for analysis. Considering that the charging process will not have any effect on the discrimination after the interface between the sample solution and the running buffer moves into the capillary, we could apply a small voltage for a certain time and then apply a larger injection voltage to improve the efficiency. We would need to know the appropriate time to change the injection voltage. What we did was to apply a linear ramped injection voltage, as shown in Figure 5A. Applying a ramped injection voltage is like applying a very small injection voltage each step. In this way, the potential of the sample solution can follow fast with the potential of the running buffer, and therefore the discrimination could be minimized. Through control of the ramped rate of the injection voltage, we can reduce the discrimination while minimizing extra band broadening caused by the injection. Figure 5B shows the simulated effects of the ramped rates on ∆V. It shows that ∆V decreases with decrease in the ramped rates. In practical analysis, we can determine which ramped rates will be most suitable through testing the area ratios of the different sample species at the different ramped rates. If the area ratios of the different sample species are nearly constant when the ramped rate of the
injection voltage is below a particular rate, that ramped rate could be considered to be the best. On-Column Fracture/Electrokinetic Injection Combined with Ramped Injection Voltage. Figure 6 shows the results obtained using the two-electrode method combined with ramped injection voltage. We can see from it that the discrimination decreased with decrease in the ramped rates, which was in agreement with the simulation results in Figure 5. Compared with Figure 2, the area ratios of the three sample components in deionized water were nearly the same as that introduced by hydrodynamic injection within experimental errors when ramped rate was 166 V/s. Further experiments demonstrated that no deviation was found when using this method to inject sample solutions with different conductivity compared to that introduced
by hydrodynamic injection. Based on the results obtained, we can conclude that this electrokinetic injection method can potentially be considered as a no-discrimination injection method suitable for sample solutions with different conductivity and will not cause severe errors in practical analysis. ACKNOWLEDGMENT We thank Mr. Yuansheng Wu for his help. We also acknowledge the National University of Singapore for financial support.
Received for review August 28, 1997. Accepted March 17, 1998. AC9709514
Analytical Chemistry, Vol. 70, No. 11, June 1, 1998
2253
