Anal. Chem. 1998, 70, 3695-3698
Technical Notes
Stacking from the Sample Stream in CZE Using a Pneumatically Driven Computerized Sampler Ruth Kuldvee and Mihkel Kaljurand*
Department of Chemistry, Tallinn Technical University, Ehitajate tee 5, Tallinn EE0026, Estonia
It is demonstrated that a pneumatically driven computerized sampling device for capillary electrophoresis facilitates sample stacking by the head-column field amplification (HCFA) technique. This device utilizes a rapid exchange between buffer and sample in a narrow channel at the separation capillary inlet and makes possible the combination of two classical injection modessthe electrokinetic and hydrodynamic modes. Detection limits obtained were about 9 nM for alkylbenzylamine cations with common UV detection. Recent developments in nonconventional sample introduction in capillary electrophoresis have focused on the possibility of forcing the sample stream to pass the separation capillary inlet. The advantages of such input devices are the absence of a voltage rise/drop time during sampling, ease of operation because no vial manipulations are involved, and ease of automation and computerization. Such an input system has been described by Verheggen et al.1 and Bushey and Jorgenson.2 Recently, Kuban and coworkers3,4 and Fang et al.5 developed an inlet system for using CE as a detector for a flow injection analysis technique. The authors of this note developed a pneumatically driven computerized sampler for CE.6 Liu and Dasgupta described an ingenious falling drop sample introduction system.7 In all these systems, the sample is believed to be introduced by the well-known electrokinetic phenomenon during the time when sample is present or passes the capillary inlet. However, since sampling requires periodic rinsing of the input channels by buffer and sample solution, sample could well be introduced hydrodynamically by pressure applied to the sample solution. In many samplers cited above, the amount of the sample introduced by the latter mechanism is probably negligible. In this note, we report that a certain amount of the sample is introduced into the capillary head during the rinse and demonstrate that this facilitates head-column stacking of sample from (1) Verheggen, Th. P. E. M.; Beckers, J. L.; Everaerts, F. M. J. Chromatogr. 1988, 452, 615. (2) Bushey, M. M.; Jorgenson, J. W. Anal. Chem. 1990, 62, 978. (3) Kuban, P.; Karlberg, B. Anal. Chem. 1997, 69, 1169. (4) Kuban, P.; Engstro¨m, A.; Olsson, J. C.; Thorse´n, G.; Tryzell, R.; Karlberg, B.; Anal. Chim. Acta 1997, 337, 117. (5) Fang, Z.-L.; Liu, Z.-S.; Shen, Q. Anal. Chim. Acta 1997, 346, 135. (6) Kaljurand, M.; Ebber, A.; Somer, T. J. High Resolut. Chromatogr. 1995, 18, 263. (7) Liu, H.; Daskupta, P. K. Anal. Chem. 1997, 69, 1211. S0003-2700(98)00111-5 CCC: $15.00 Published on Web 07/03/1998
© 1998 American Chemical Society
the stream flowing past the capillary inlet if the sample solution conductivity is much lower than the running buffer conductivity. Head-column stacking was demonstrated to be a powerful and simple way to reduce detection limits in CE.8-11 Analyte detection limits can be improved by more than 2 orders of magnitude compared to the common electrokinetic sampling. In this work, the detection limits achieved for two cations were 250 ppb for the latter method, when for the first one it was 2.5 ppb (9 nM) using common UV detection. EXPERIMENTAL SECTION Equipment. The CE system consisted of a homemade autosampler, a homemade high-voltage supply delivering 18 kV, an 80-cm full length (50 cm to detector), 50-µm-i.d. capillary (Polymicro Technologies, Phoenix, AZ), and an Isco CV4 UV detector. Detector signal was digitized and transferred to the 486type computer via a Keithley ADC-16 analog-to-digital board. The same board delivered digital signals to the solenoid valves controlling the autosampler. The autosampler schematics is shown in Figure 1. It operates on the principle of rapidly changing between buffer and sample in the input channel of the autosampler block by pressure applied either to the buffer or to the sampler vessel. Total dimensions of the plexiglass sampler body are 35 × 25 × 25 mm3. Input channel has length of 45 mm, with 1.5 mm i.d. Thus, the total input channel volume is about 80 µL. Assuming that the input channel and capillary are filled with running buffer and high voltage is applied, the sampling is performed according to the sequence of pressures applied as shown in Table 1. It is evident from Table 1 and Figure 1 that, to execute the sampling process, pressures must satisfy the following relationship: P1 ) P2 < P3 ) P4. The pressures are delivered by two solenoid valves. A thorough description of the sampler is given in ref 6. Chemicals. Each analysis was performed with sodium phosphate buffer as background electrolyte at pH ) 6.7 and ionic strength I ) 0.1. All solutions were prepared by dissolving chemicals of analytical grade in deionized water. Benzyltriethylammonium chloride (BTEA) and benzyltributylammonium chloride (BTBA), both purchased from Merck, were used as samples. (8) (9) (10) (11)
Chien, R. L.; Burgi, D. S. Anal. Chem. 1992, 64, 489A. Burgi, D. S. Anal. Chem. 1993, 65, 3726. Martinez, D.; Borrull, F.; Calull, M. J. Chromatogr. A 1997, 788, 185. Zhang, C.-X.; Thormann, W. Anal. Chem. 1996, 68, 2523.
Analytical Chemistry, Vol. 70, No. 17, September 1, 1998 3695
Figure 2. Illustration of the effect of stacking from the sample stream. (a) Rinse time 2.5 s, sample kept in input channel for 0.25 s, high voltage switched off; (b) rinse time 2.5 s, electrokinetic sampling time 0.25 s; (c) rinse time 0.25 s, electrokinetic sampling time 2.5 s. Peaks: 1, BTEA; 2, BTBA; 3, water; ?, unknown. Note the differences in sample concentrations: in a, b, and c, the sample concentrations were 125, 15, and 250 ppb, correspondingly.
Figure 1. Pneumatic autosampler: (A) sample rinse, (B) buffer rinse. Table 1. Sampling Sequence Logic step
action
P1
P2
P3
P4
duration
1 2 3 4
sample rinse electrokinetic sampling buffer rinse pherogram run
off off on off
on off off off
on on off on
off on on on
0.125-10 s 0.25-10 s 0.125-0.25 s 5 min-1 h
Sodium hydroxide was obtained from Chempol and phosphoric acid from YA-Kemia. Solutions were filtered through 0.45-µm Millipore filters. RESULTS AND DISCUSSION It follows from the description of the operation of the sampler that the time spent in step 2 determines the electrokinetic sampling time of the autosampler and that steps 1 and 3 should be as short as possible to enable pure electrokinetic sampling. Optimization of the operating parameters of the autosampler has been studied and will be published in a separate paper. Only the main results will be given here. The pressure actuating the buffer/sample stream should be within (0.3-0.85) × 105 Pa. If the pressure is outside these limits, severe reduction of the separation efficiency results, together with large baseline irregularities. This is probably due to overloading of the capillary with sample if the pressure is higher than 0.85 × 105 Pa; sampleto-buffer exchange does not occur with the required speed if the pressure is lower than 0.3 × 105 Pa. This pressure is determined by the length and diameter of the sampler channels and connecting tubing, thus being specific for the particular sampler design. 3696 Analytical Chemistry, Vol. 70, No. 17, September 1, 1998
If the pressure is within the limits, sample-to-buffer (or vice versa) replacement (or rinse time) requires about 0.125 s. The separation efficiency (expressed, e.g., in theoretical plate numbers) decreases with increasing rinse time. But if the sample concentration is low enough (1 ppm and lower) and the sample is dissolved in water, then it is possible to increase the rinse time to 10 s without serious loss of efficiency. In this case, sample stacking could be expected by the field-amplified effect. The situation is illustrated in Figure 2, where three different sample introduction techniques have been compared. Different concentrations for comparisons were used deliberately, to have measurable sample component peaks at least in two of three electropherograms. First, the sample was introduced hydrodynamically only by forcing the sample stream to flow through the input channel while the high voltage was switched off. The duration of steps 1 and 3 was 2.5 s, and the duration of step 2 was 0.25 s. After sampling, the high voltage was switched on again. In this case, the sample was introduced only by pressure applied to the sample stream. The electropherogram (Figure 2a) consists of a water peak only; no BTEA and BTBA peaks can be detected since their concentrations are only about 125 ppb. The same procedure was repeated with the same durations for sampling steps 1-3, with high voltage now applied. The electropherogram (Figure 2b) now consists of a water peak with the same size as in the former case plus two cation peaks present with very good signalto-noise ratio, despite the fact that the concentrations of the analytes are much lower (15 ppb). The fact that water peaks are the same size in both cases demonstrates that the sample is introduced, indeed, hydrodynamically during the rinse time and the contribution of electroosmosis is negligible. In the first experiment, with high voltage switched off, the stacking occurs only within the sample zone introduced into the column. In the second case, when high voltage is switched on during the sampling, sample probably stacks to the column head from the whole volume around the capillary inlet. A third experiment was performed using short (0.25 s) rinse time (steps 1 and 3) but with
Figure 3. Electropherogram of BTEA and BTBA. Sample concentration 2.5 ppb. Peaks: 1, BTEA; 2, BTBA; 3, water.
sample in the input channel for 2.5 s, and regular electrokinetic sampling took place. The electropherogram has much smaller water peak than in the previous two cases and two low-intensity cation peaks, despite the fact that the total duration of the sampling steps 1 and 2 is equal in the second and third experiments. Concentrations of the analytes are 250 ppb. The relationship between peak area and rinse time is linear up to the rinse time of 10 s, with a correlation coefficient of 0.9989. The reproducibility of peak areas is about 2.3% if the sample is stacked from the sample stream. If the rinse time is longer than 10 s, the current through the capillary drops significantly (by 95%) during the rinse time, and the corresponding electropherogram has an irregular baseline with deteriorated peak shapes. Figure 3 presents an electropherogram of a 2.5 ppb sample (about 9 nM of both analytes), thus exhibiting the detection limit achieved in present work. If we compare this detection limit with those reached with electrokinetic injections250 ppb (see Figure 2)s the detection limit is 2 orders lower in the first case. Thus, a sample plug can be introduced to the column head by three different means: (1) from the flowing sample stream by pressure with high voltage applied simultaneously, (2) from the flowing sample stream by pressure with high voltage switched off, and (3) by electroosmosis during electrokinetic sampling. The mechanism responsible for the improved detection limits obtained in the first case compared to other cases is probably the following: by applying pressure, a small sample plug is applied to the capillary head, and, if high voltage is applied as well, stacking is likely to occur from the whole volume of the input channel. In pure hydrodynamic sampling, the sample plug is also introduced at the capillary head; however, since the next sampling step is to replace the capillary inlet channel with buffer, the sample amount in this plug is too small to be stacked and forms an observable peak on the electropherogram. If the sample solution is introduced electroosmotically, stacking also applies to the whole sample channel; however, the electroosmotically introduced lowconductivity plug at the column head is too small to ensure an observable stacked amount of sample in the plug.
Figure 4. Electropherogram of the BTEA and BTBA. Sample concentration 10 ppb. Rinse time for electropherograms a and b was 2.5 s, electrokinetic sampling time was 2.5 and 5.0 s, accordingly. Rinse time for electropherograms c and d was 5 s, electrokinetic sampling time was 5 and 10 s. (a,c) High voltage switched on during rinse time; (b,d) high voltage switched off during rinse time. Peaks: 1, BTEA; 2, BTBA.
To understand the mechanism of the injections with long rinse more clearly, we studied the sampling in two cases: either the sample is flowing past the capillary inlet or standing still. In the context of this paper, it would be convenient to call the stacking from the stagnant sample static stacking and the stacking from the sample stream dynamic stacking. To establish the possible difference between the two techniques, the following two experiments were performed. In the first experiment, sampling consisted of a long rinse time (step 1 in Table 1), and electrokinetic sampling time (step 2 in Table 1) with the two steps having equal duration. High voltage was on during the whole sampling procedure. In the second experiment, rinse time was equal to that in the first experiment, but electrokinetic sampling time was twice as long as in the first experiment. In the second experiment, high voltage was switched off during the rinse time. In the second case, the contribution of the pure hydrodynamic sample introduction remains, but if the concentration is low the contribution can be ignored. If the stacking from the flow does not involve any peculiarity compared to the static stacking, the peak areas of cations must be the same in the first and the second experiments when the high voltage was switched off or on because the head-column HCFA phenomenon applies in both cases. At a sample concentration of 50 ppb, the cations peak area ratio was about 2 times between dynamic and static stacking, and the increase in the rinse time did not influence the results significantly. At a lower sample concentration (10 ppb), increasing the rinse time by a factor of 2 increased further the peak areas ratio approximately 2.5 times. Figure 4 illustrates the results. A possible explanation is that, in the case of long static sampling, the sample channel can be drained empty of cations: i.e., the predominant part of the sample was expelled and sample stacking took place only from the sample portion remaining in the input channel, which was rapidly drained off the sample. Independent studies by Zhang and Thormann also indicate that input vessels can be drained empty if the HCFA Analytical Chemistry, Vol. 70, No. 17, September 1, 1998
3697
occurs.12 In experiment 1, where high voltage was on during the sample rinse, the total sample volume from which the stacking occurs is larger (due to the flow of fresh sample into input channel) than in experiment 2. Thus, this experiment confirms our hypothesis that, during rinse time, a water plug penetrates the capillary and ensures head-column field amplification and that there exist some differences between dynamic and static stacking. Such differences increase with decreasing concentration. To give
a satisfactory explanation needs further studies. Another advantage of the dynamic stacking should be mentioned. In ref 11, the authors speculate about possible electrolysis degradation and overheating of the sample. Stacking from the stream should avoid such disadvantages.
(12) Zhang, C.-X.; Thormann, W. Anal. Chem. 1998, 70, 540.
AC9801115
3698 Analytical Chemistry, Vol. 70, No. 17, September 1, 1998
Received for review February 2, 1998. Accepted May 22, 1998.
