Anal. Chem. 1999, 71, 1638-1644
Sweeping of Analyte Zones in Electrokinetic Chromatography Joselito P. Quirino* and Shigeru Terabe
Faculty of Science, Himeji Institute of Technology, Kamigori, Hyogo, Japan 678-1297
The sweeping phenomenon in electrokinetic chromatography under a considerably homogeneous electric field is theoretically examined and experimentally verified. The roles of analyte charge, retention factor, and electroosmotic flow are given consideration. Experimental studies are performed using micellar electrokinetic chromatography with sodium dodecyl sulfate as a micelle-forming agent. Uses and limitations of the new phenomenon are also critically conveyed. Electrokinetic chromatography (EKC) is a mode of capillary electrophoresis, which effectively separates neutral substances.1,2 EKC also provides added selectivity in the separation of ionic substances that are sometimes difficult to separate by capillary zone electrophoresis (CZE).3,4 The separation mechanism involves partitioning of analytes between the pseudostationary phase and the surrounding aqueous phase and electrokinetic phenomenon. Since its introduction in 1984, using micelles as pseudostationary phases (preferably termed as MEKC), significant advances in theory and novel applications have been documented. However, some questions still need answers including effects of sample matrix,5,6 solute adsorption into the capillary wall, broadening mechanisms, among other things. In our search for ways to improve concentration sensitivity of neutral analytes on capillary by sample stacking in MEKC,7-12 we discovered a new phenomenon now termed as sweeping.13 Sweeping is the picking and accumulation of analyte molecules by the pseudostationary phase that penetrates the sample zone. This causes a unique focusing effect. The injected length of a neutral analyte zone was found to be theoretically narrowed by a factor equal to 1/(1 + k) (k, retention factor) and the concentration can be increased approximately by a factor, 1 + k. These were * Corresponding author: (tel) +81-791580173; (fax) +81-791580132; (e-mail)
[email protected]. (1) Terabe, S.; Otsuka, K.; Ichikawa, K.; Tsuchiya, A.; Ando, T. Anal. Chem. 1984, 56, 111-3. (2) Terabe, S. Trends Anal. Chem. 1989, 8, 129-34. (3) Otsuka, K.; Terabe, S.; Ando, T. J. Chromatogr. 1985, 348, 39-47. (4) Quang, C.; Strasters, J. K.; Khaledi, M. G. Anal. Chem. 1994, 66, 164653. (5) Shihabi, Z. K.; Hinsdale, M. E. J. Chromatogr., B 1995, 669, 75-83. (6) Shao, L. K.; Locke D. C. Anal. Chem. 1998, 70, 897-906. (7) Quirino, J. P.; Terabe, S. J. Chromatogr., A 1997, 781, 119-28. (8) Quirino, J. P.; Terabe, S. J. Chromatogr., A 1997, 791, 255-67. (9) Quirino, J. P.; Terabe, S. Anal. Chem. 1998, 70, 149-57. (10) Quirino, J. P.; Terabe, S. J. Chromatogr., A 1998, 798, 251-7. (11) Quirino, J. P.; Terabe, S. Anal. Chem. 1998, 70, 1893-901. (12) Quirino, J. P.; Otsuka, K.; Terabe, S. J. Chromatogr., B 1998, 714, 29-38. (13) Quirino, J. P.; Terabe, S. Science 1998, 282, 465-468.
1638 Analytical Chemistry, Vol. 71, No. 8, April 15, 1999
under conditions of relatively constant electric fields throughout the capillary, negligible electroosmotic flow, and neutral analytes prepared in a matrix void of the pseudostationary phase. When k goes to infinity, only the column length to the detector restricts injection lengths that will provide an effective separation length window. Nevertheless, the major physical impact was that high-k analytes could be concentrated without bounds and separation efficiencies preserved or enhanced. Detection sensitivity of some analytes was improved 5000-fold without off-line treatment, by MEKC with UV detection. A racemic herbicide spiked in lake water was also separated and detected at ∼15-ppb level with very minimal sample treatment (the conductivity was only adjusted to that of the separation solution). Furthermore, understanding on the fate of analyte zones in EKC was broadened. In this article, detailed discussion on the sweeping phenomenon is given. The study incorporates the role of analyte charge and electroosmotic flow aside from the retention factor. Theoretical models were developed and experimentally studied. Moreover, limitations and practical aspects of sweeping are discussed. THEORY Length of the Bands after Sweeping under a Homogeneous Electric Field. (a) Neutral Molecules. Figure 1 depicts our theoretical model for the sweeping of neutral analytes (a) in the presence of electroosmotic flow. Movement of all pertinent boundaries upon application of voltage is illustrated. All electrokinetic velocities or electrophoretic mobilities are positive and negative when movement is directed toward the cathode and anode, respectively. In the starting situation (Figure 1A), injection of sample solution (S) having the same conductivity as the background solution (BGS), ac and aa are the neutral analyte molecules found near the interface between S and BGS zones at the cathodic and anodic ends, respectively. The mcc and mca are the micelles at the cathodic and anodic ends, correspondingly. All corresponding electrokinetic velocities are assumed similar in both the S and BGS zones, unless stated. The length of the S zone injected is given as linj. The arrows in Figure 1A show the magnitude of the electrophoretic mobilities. The broken lines between parts A-C indicate the starting position of ac and mcc. Upon application of voltage (Figure 1B), micelles at the cathodic side that migrate toward the cathode will sweep a (shaded area). The ac will be incorporated into the micelle; thus the migration velocity is equal to that in the usual MEKC or higher than that of the micelle. However, the migration velocity of aa is equal to the electroosmotic flow (veof) because it will not be incorporated into any micelle until the time when aa reaches mcc 10.1021/ac9810866 CCC: $18.00
© 1999 American Chemical Society Published on Web 03/16/1999
Figure 1. Evolution of micelles and neutral analyte molecules during sweeping in the presence of high electroosmotic flow. (A) Starting situation, injection of S prepared in a matrix having a conductivity similar to that of the BGS; (B) application of voltage at positive polarity, micelles emanating from the cathodic side sweeping analyte molecules; (C) the injected analyte zone is assumed completely swept. Other symbols and explanations in the text.
Figure 2. Evolution of micelles and negatively charged analyte molecules during sweeping in the presence of high electroosmotic flow where the electrophoretic mobility of the analyte is greater compared to the micelle. (A) Starting situation, injection of S prepared in a matrix having a conductivity similar to that of the BGS; (B) application of voltage at positive polarity, micelles emanating from the anodic side sweeping analyte molecules; (C) the injected analyte zone is assumed completely swept. Other symbols and explanations in the text.
from the cathodic end. The length of the a zone after sweeping (lsweep) is then given as eq 1 (see Figure 1C). This is assumed to
lsweep ) d(ac) - d(mcc)
(1)
be when aa reaches mcc. The distances traveled by mcc [d(mcc)] and ac [d(ac)] are given by eqs 2 and 3, respectively, where vmc is
d(mcc) ) vmctsweep; vmc ) vep(mc) + veof d(ac) ) va(MEKC)tsweep; va(MEKC) ) vep*(a) + veof
(2) (3)
the migration velocity of the micelle, va(MEKC) is the migration velocity of ac, vep(mc) is the electrophoretic velocity of the micelle (eq 4), vep*(a) is the effective electrophoretic velocity of a (eq 5), and tsweep is the time when aa reaches mcc (eq 6). Also, veof is given by eq 7.
vep(mc) ) µep(mc)E
(4)
vep*(a) ) (k/(1 + k)) µep(mc)E
(5)
tsweep ) linj/(veof - vmc)
(6)
veof ) µeofE
(7)
where µep(mc) is the electrophoretic mobility of the micelle, E is the electric field strength, k is the retention factor, and µeof is the coefficient of electroosmotic flow or electroosmotic mobility. Algebraic manipulation of all equations above yields the final equation for lsweep (eq 8). Generally, eq 8 implies that the length
lsweep ) linj(1/(1 + k))
higher retention factor analytes. Second, an increase in linj will increase the length of the swept zones for each analyte. Finally, electroosmotic flow has no effect on the length of the swept zones. (b) Charged Molecules. Figure 2 depicts our theoretical model for the sweeping of a negatively charged analyte (a′) in the presence of electroosmotic flow. The symbols and other assumptions are similar to those given in Figure 1, except the charged analyte is given as a′. The broken lines between parts A-C of Figure 2 indicate the starting position of a′a and mca. Note that since the analyte is charged it has electrophoretic mobility, and in this example, the magnitude is greater compared to the micelle. Upon application of voltage (Figure 2B), micelles at the anodic side will sweep the slower moving a′ near the anodic side (shaded area). The migration velocity of a′a follows that in MEKC for charged analytes. On one hand, the migration velocity of a′c is equal to that in the usual CZE. This is because it will not be incorporated into any micelle until the time when micelles from the anodic end reach it. The length of the zone of a charged analyte after sweeping (l′sweep) is then given by eq 9 (see Figure
l′sweep ) d(mca) - d(a′a)
(9)
2C). Similarly, this is assumed to be when mca reach a′c. The distance traveled by mca [d(mca)] is equal to d(mcc) (eq 2). The distance traveled by a′a [d(a′a)] is given by eq 10, where va′(MEKC)
d(a′a) ) va′(MEKC)tsweep; va′(MEKC) ) vep*(a′) + veof
(8)
(10)
of the sample zones after sweeping given a fixed linj is shorter for
is the migration velocity of a′a and vep*(a′) is the effective Analytical Chemistry, Vol. 71, No. 8, April 15, 1999
1639
electrophoretic velocity of a′ (eq 11)14 where µep(a′) is the
vep*(a′) )
[1 +1 k µ
ep(a′)
+
]
k µ (mc) E 1 + k ep
(11)
electrophoretic mobility of a′ and k is given by eq 12.3,4 The µ(a′)
k ) (µ(a′) - µep(a′))/(µep(mc) - µ(a′))
(12)
is the electrophoretic mobility of a′ in the MEKC system. The tsweep in eq 10 is the time when mca reaches a′c (eq 13), where
tsweep )
linj vmc - va′(CZE)
(13)
va′(CZE) is the migration velocity of a′ in CZE or the absence of surfactant (eq 14).
va′(CZE) ) {µep(a′) + µeof}E
(14)
Substitution of eqs 2, 10, 11, 13, and 14 to eq 9 and algebraic manipulation yields the final equation for l′sweep (eq 15). The same
l′sweep ) linj(1/(1 + k))
(15)
equation was derived when the electrophoretic mobility of the micelle is greater than that of the analyte. The same equation above was also derived for positive analytes, whether the electrophoretic mobility of the micelle is greater than that of the analyte and vice versa. Upon further examination, the above equation is similar to that derived for neutral analytes except the k (eq 12) is computed differently. It should also be emphasized that the retention factor plays the major role in the narrowing of ionic analyte zones and the practical applicability of sweeping will be for ionic analytes having great affinities toward the pseudostationary phase. With this in mind, anionic analytes that have low k values when using anionic micelles as the pseudostationary phase will not be concentrated effectively. EXPERIMENTAL SECTION Apparatus. All capillary electropherograms were obtained with a Hewlett-Packard 3D capillary electrophoresis system (Waldbronn, Germany). Electrophoresis and sweeping experiments were performed in fused silica capillaries of 50 µm i.d. and 375 µm o.d. obtained from Polymicro Technologies (Phoenix, AZ). A neutral capillary (CElectTM-N) of 50 µm i.d. and 360 µm o.d. obtained from Supelco (Bellefonte, PA) was also used. Capillaries were thermostated at 20 °C. Wavelengths of detection for each analyte were selected using spectral absorbance curves recorded with a diode array detector. Water was purified with a Milli-Q system (Millipore, Bedford, MA). Conductivities were measured using a Horiba ES-12 conductivity meter (Kyoto, Japan). Reagents and Solutions. All reagents were purchased in the highest grade possible from Nacalai Tesque (Kyoto, Japan). Stock solutions of 0.5 M sodium dodecyl sulfate (SDS) were prepared (14) Vindevogel J.; Sandra, P. Introduction to Micellar Electrokinetic Chromatography; Huthig: Heidelberg, 1992.
1640 Analytical Chemistry, Vol. 71, No. 8, April 15, 1999
every week in purified water. Micellar BGS were prepared (each day) by dilution of the SDS stock solution in appropriate phosphate or borate buffers. Stock solutions of alkyl phenyl ketones (decanophenone, hexanophenone, butyrophenone, propiophenone, acetophenone), aromatic amines (1-naphthylamine, 1-phenylethylamine, benzylamine, aniline), steroid derivatives (progesterone, testosterone, hydrocortisone, cortisone), phenols (2,4-dichlorophenol, pentachlorophenol), and dibutyl phthalate were prepared in methanol. Stock solutions of salicylic and benzoic acids were prepared in 50% methanol. Concentrations of alkyl phenyl ketones, aromatic amines, and dibutyl phthalate stock solutions ranged between 5200 and 6100 ppm. Concentrations of phenols were 1310 and 4850 ppm for 2,4-dichlorophenol and pentachlorophenol, respectively. Concentration of salicylic and benzoic acids were 6500 and 6000 ppm, respectively. Polycyclic aromatic hydrocarbons (PAHs, anthracene, phenanthrene, fluorene, acenaphthene, biphenyl, acenaphthylene, naphthalene) were prepared in acetonitrile to concentrations ranging from 1000 to 2000 ppm. Care should be taken when handling PAHs and other aromatic compounds listed above as they may be disease-causing agents. Portions of the appropriate stock solutions were combined and diluted appropriately in the BGS (or 1:1 ratio of BGS/acetonitrile in the case of the PAHs) to determine concentrations wherein analytes possess comparable peak heights. Secondary stock solutions having analyte concentrations between 100 and 200 ppm were then prepared in appropriate phosphate or borate buffer solutions (with some amounts of Brij 35 or organic solvent as indicated in the figures). Brij 35 is chemically known as polyoxyethylene (23) lauryl ether. The sample matrixes were previously adjusted to the conductivity of the BGS by titrating with a higher or lower concentration of buffer solution. Final dilutions (sample solution, S, concentrations in the figures) were done using the same buffers above. All solutions were filtered through 0.45 µm filters (Toyo Roshi, Japan) prior to capillary electrophoresis experiments. General Electrophoresis Procedure. The capillary was conditioned prior to use with 1 M NaOH (10 min), followed by methanol (5 min), purified water (5 min), and finally BGS (5 min). The S was then injected into the capillary farthest from the detector end using pressure (50 mbar). The velocities of a liquid at 50- or ∼1000-mbar pressure were determined to approximate the lengths of the zones injected at different intervals. Voltage was then applied at negative (low-pH buffers) or positive polarity (high-pH buffers) with the BGS at both sides of the capillary, until all peaks are detected. The capillary was flushed, between consecutive analyses to ensure repeatability, with methanol (1 min), followed by 0.1 M NaOH (1 min), purified water (2 min), and finally BGS (2-5 min). Other conditions are specified in the text or figures. RESULTS AND DISCUSSION Effect of Electroosmotic Flow. To show the independence of sweeping from the electroosmotic flow, neutral analyte widths under sweeping conditions obtained with different electroosmotic flow velocities were measured. This is summarized in Figure 3, which are plots of lsweep versus linj for each k. Acetophenone (k ) 1.64), propiophenone (k ) 3.53), and butyrophenone (k ) 8.36) were used as test analytes. The k values were computed using
Figure 3. Theoretical and experimental lsweep versus linj for the test alkyl phenyl ketones in the absence and presence of electroosmotic flow. (A) negligible electroosmotic flow; (B) low electroosmotic flow; (C) high electroosmotic flow. Theoretical plots in each plate (lines): 1, acetophenone; 2, propiophenone; 3, butyrophenone. Experimental values in each plate (symbols): (triangles) acetophenone; (squares) propiophenone; (circles) butyrophenone. BGS, 50 mM SDS in 50 mM phosphate buffer (A, B) or 20 mM borate buffer (C); S, ∼5 ppm of each alkyl phenyl ketone in phosphate or borate buffer with almost the same conductivity and pH compared to that of the corresponding BGS; capillary, 56 cm to the detector (64.5 cm total); more information in the text. Table 1. Experimental Peak Widths (cm) and Plate Numbers (×105) Obtained from a 1-s Injection no EOFa
low EOFb
Table 2. Theoretically and Experimentally Determined l′sweep Values
high EOFc
width plate no. width plate no. width plate no. acetophenone propiophenone butyrophenone
0.39 0.35 0.37
1.5 1.9 1.7
0.74 0.85 0.97
0.4 0.3 0.3
0.66 0.62 0.60
0.5 0.6 0.6
a Other conditions are the same as those found in Figure 3A. b Other conditions are the same as those found in Figure 3B. c Other conditions are the same as those found in Figure 3C.
the conditions in Figure 3C with decanophenone as a marker of the micelle. Methanol was used as a marker of the electroosmotic flow or decanophenone in the absence of surfactant, with no significant difference observed. Phosphate (A, B) or borate (C) buffers having different pH values were used to control the electroosmotic flow velocity. The pH values were 1.9 (A), 4.5 (B), and 9.3 (C), corresponding to negligible, low, and high electroosmotic flows, respectively. The lines are plots obtained from theory (eq 8), and the symbols are experimental values. The plots obtained experimentally for all the cases are somewhat similar to the theoretical plot, as expected in the Theory section. However, if the experimental values in all cases are linearly regressed, plots for (B) and (C) do not pass the origin nicely compared to (A). This could be explained by a former observation in normal MEKC injections, where peak widths obtained in systems with electroosmotic flow are broader compared to those under zero electroosmotic flow (see Table 1), which is still not fully understood.15 Moreover, experimental lsweep values in (B) are more problematic, and the probable reason is the irreproducible electroosmotic flow at the studied pH, causing incorrect observed values. The relative (15) Janini, G. M.; Issaq, H. J.; Muschik, G. M. J. Chromatogr., A 1998, 792, 125-41.
A. Positively Chargeable Compounds under Zero Electroosmotic Flowa l′sweep
1. 1-naphthylamine 2. 1-phenylethylamine 3. benzylamine 4. aniline
kb
theoryc
exp
7.5 4.4 3.3 1.9
0.35 0.56 0.69 1.04
0.20 0.35 0.53 0.54
B. Negatively Chargeable Compounds under High Electroosmotic Flow 1. benzoic acid 0.06 2.83 2. salicylic acid 0.08 2.78 3. pentachlorophenol 0.45 2.07
2.05 2.03 1.23
a Conditions: BGS, 25 mM SDS in 50 mM phosphoric acid (pH 1.9) (A), 50 mM SDS in 20 mM sodium dihydrogen phosphate (pH 8.75) (B); linj 3 cm; concentration of analytes, ∼15 ppm (A), 10 ppm (B); applied voltage, -20 kV (A), 22 kV (B); capillary, 56 cm to the detector (64.5 cm total). b Computations based on eq 12, where µep(a′) is determined using the conditions above in the absence of SDS in the BGS (B) and at positive polarity (A). c Calculatations based on eq 15.
standard deviation of the electroosmotic flow velocity at pH 4.5 is 11%. Use of pH 3.5 and 5.5 also incurred similar undesirable results. The charge at the surface of the capillary is not homogeneous at pH between 3 and 5.5 due to differences in the number of protonated silanol groups. Nonetheless, the trend of eq 8 is unambiguously confirmed experimentally. Effect of Analyte Charge. Table 2 lists the theoretical and experimental values of lsweep for positively charged analytes under zero electrosomotic flow (A) and negatively charged analytes under high electroosmotic flow (B). These were representative experiments done to confirm eq 15. The linj was kept at 3 cm for all experiments. The theoretical and experimental values in Table 2 are quite comparable, confirming the sweeping equation for Analytical Chemistry, Vol. 71, No. 8, April 15, 1999
1641
Figure 4. Sweeping MEKC analysis of some hydrophobic test analytes in the presence of small amount of organic solvent in the sample matrix: BGS, 50 mM SDS in 50 mM phosphoric acid (pH 1.9)/20% acetonitrile; S, test analytes in 2% acetonitrile in 50 mM phosphoric acid adjusted to the conductivity of the BGS (5.56 mS/cm); injected length of S, 0.64 mm (A), 64 cm (B); applied voltage, -30 kV; concentration of analytes, ∼300 ppm (A), ∼0.3 ppm (B); identification of peaks, dibutyl phthalate (1), hexanophenone (2), naphthalene (3); capillary, 72 cm to the detector (80.5 cm total).
charged analytes in the absence or presence of electroosmotic flow. Practical Considerations of the Sweeping Phenomenon. (a) Nature of Analyte. As shown in a previous paper, at least 2000-fold improvements in detector response were obtained with some hydrophobic neutral analytes (steroids, phenols) using the sweeping phenomenon.13 Enhancement factors were calculated by simply getting the ratio of the peak heights obtained from sweeping and normal injection (1- or 2-s injection of S) and correction by the dilution factor. The problem with the hydrophobic neutral analytes is their low solubility in aqueous media; however, most of these analytes are soluble in the ppb-level range in aqueous systems, which makes them good samples for the technique. Direct injection had been demonstrated. Addition of organic solvent is recommended for extremely hydrophobic analytes; however, enhancement factors will be reduced depending on the amount of solvent added. The k in the S zone, which is the most important parameter in sweeping, decreases with the increase in the concentration of organic solvent. Figure 4 shows that ∼300-fold to more than 1000-fold improvement in detection sensitivity can be obtained for some organic neutral compounds even with the addition of a small amount of acetonitrile to the sample matrix. This may be explained by the still reasonably high k with only 2% acetonitrile in the sample matrix. Addition of 20% acetonitrile to the BGS does not affect sweeping because it has no effect on the k values in the S zone. A usual injection (Figure 4A) was included for comparison. Note that the S in Figure 4B is a 1000-fold dilution of the S in Figure 4A. Spikes in the sweeping electropherogram (Figure 4B) are caused by the changes in composition of the liquid during sweeping.13 The change in migration time, especially for the naphthalene peak, is caused by the difference in k between the S and BGS zones. Each test analyte was successfully identified using a recorded UV spectrum. Addition of nonionic surfactant into the matrix may be a feasible alternative to improve solubility of hydrophobic neutral 1642 Analytical Chemistry, Vol. 71, No. 8, April 15, 1999
Figure 5. Sweeping MEKC analysis of seven PAHs in the presence of Brij 35 in the sample matrix and separation solution: BGS, 50 mM SDS, 0.02 mM Brij 35, and 35% methanol prepared in 100 mM phosphoric acid (pH 1.9); S, PAHs in 0.05 mM Brij in 50 mM phosphoric acid adjusted to the conductivity of the BGS (pH 1.9); injected length of S, 1.3 mm (A), 10.3 cm (B); applied voltage, -25 kV; concentration of analytes, ∼70 ppm (A), ∼0.7 ppm (B); identification of peaks, anthracene (1), phenanthrene (2), fluorene (3), acenaphthene (4), biphenyl (5), acenaphthylene (6), naphthalene (7); capillary, 56 cm to the detector (64.5 cm total).
analyes in the sample matrix. A small concentration of the nonionic surfactant should be added into the BGS to minimize baseline noise caused by the nonionic surfactant when it reaches the detector. Remember that nonionic surfactants form mixed micelles with charged ones. Preliminary work was performed with very difficult samples (PAHs), wherein the concentration in solution at low levels is very erratic due to adsorption on containers and capillary walls. Brij 35 was used to stabilize the sample solution. Adsorption is still present as illustrated in Figure 5 (only 20300-fold improvements were achieved), and reproducibility was rather poor (also with the normal injection procedure, Figure 5A). Note that the S in Figure 5B is a 100-fold dilution of the S in Figure
Figure 6. Sweeping MEKC analysis of two positively chargeable compounds. BGS, 75 mM SDS and 10 mM triethanolamine in 20% methanol and 30 mM phosphoric acid (pH 2.4); S, amines in phosphoric acid adjusted to the conductivity of the BGS; injected length, 0.64 mm (A), 32 cm (B); applied voltage, -25 kV; concentration of analytes, ∼300 ppm (A), 0.3 ppm (B); identification of peaks, 1-naphthylamine (1), 1-phenylethylamine (2); capillary, 56.5 cm to the detector (65 cm total).
5A. Separation efficiency was rather preserved considering the length of the injected S zone. Also, higher concentrations of Brij 35 in the S matrix produced very noisy electropherograms. More studies are evidently needed (e.g., use of other nonionic surfactants). The applicability to basic and acidic hydrophobic analytes is very promising. Electrostatic interactions (e.g., with cationic analytes and anionic micelles) that increase affinity of analytes to the pseudostationary phase may be tapped. Enhancement factors are greater than those obtainable by sample stacking. More than 5000-fold improvement in detector response of some weakly basic and hydrophobic drugs had been shown.13 The solubility of the analyte can also be improved by the proper choice of pH (e.g., acidic conditions for basic analytes), thus minimizing the uncertainty in determining the exact concentration of analyte. Also, chargeable hydrophobic analytes are difficult to separate by CZE alone. Figure 6 shows the several hundredfold increases in peak heights of 1-naphthylamine (∼500-fold) and 1-phenylethylamine (∼400-fold). The S in Figure 6B is a 1000-fold dilution of the S in Figure 6A. Note that, in this experiment, the solubility was improved by using a low-pH sample matrix that produced positively charged test amines. Electrostatic interaction of positively charged amines and negatively charged SDS micelles and the use of a higher concentration of SDS (compared to that used in Table 2A) provided high k values. This resulted in the better enhancement factors obtained. (b) Electroosmotic Flow. The practical utility of the sweeping phenomenon is limited to solutes of high hydrophobicity. With this in mind, the MEKC condition where electroosmotic flow is absent seems to be the best choice when sweeping of such analytes is performed. Migration times are very much shorter, plate numbers are better, and separation selectivity is easier to establish compared to when electroosmotic flow is present. Using some steroids as test analytes, detection sensitivity enhancement factors obtained without electroosmotic flow (>2000-fold for progesterone, ref 13) are better than those obtained with electroosmotic flow (∼200-fold, Figure 7). Figure 7A is a normal injection MEKC analysis, and Figure 7B is the present optimized
Figure 7. Sweeping MEKC analysis of some steroid derivatives in the presence of high electroosmotic flow: BGS, 45 mM SDS and 36 mM γ-cyclodextrin in 36 mM phosphate buffer (pH 9)/10% methanol; S, steroids in phosphate buffer (pH 9) adjusted to the conductivity of the BGS (4.0 mS/cm); injected length, 0.64 mm (A), 32 cm (B); applied voltage, 30 kV; concentration of analytes, ∼70 ppm (A), ∼0.07 ppm (B); identification of peaks, cortisone (1), testosterone (2), progesterone (3); capillary, 72 cm to the detector (80.5 cm total).
Figure 8. Sweeping with a neutral capillary and a basic buffer: BGS, 50 mM SDS and 20% methanol in 20 mM borate buffer (pH 9.3); S, steroids in borate buffer adjusted to the conductivity of the BGS; injected length, 0.64 mm (A), 42 cm (B); applied voltage, -30 kV; concentration of analytes, ∼20 ppm (A), ∼2 ppm (B); identification of peaks, progesterone (1), testosterone (2), hydrocortisone (3), cortisone (4); capillary, CelectTM-N, 57 cm to the detector (65.5 cm total).
sweeping MEKC electropherogram in the presence of high electroosmotic flow. Also, plate numbers obtained are not so good in Figure 7B. Moreover, addition of γ-cyclodextrin was essential to provide separation and to reduce migration time owing to the decrease in k. This was done because addition of organic solvents decreased the electroosmotic flow that caused very long migration times, which are undesirable. Coated capillaries and basic (or neutral) buffers for neutral (or acidic) and hydrophobic analytes is also expected to be of value. This is illustrated in Figure 8. A coated capillary (commercially available neutral capillary with negligible electroosmotic flow) and a basic micellar background solution were used to analyze several hydrophobic steroids providing almost 1000-fold increases in detection sensitivity. Note that the S in Figure 8B is a 10-fold dilution of the S in Figure 8A. Analytical Chemistry, Vol. 71, No. 8, April 15, 1999
1643
CONCLUSION In the present study, we have shown that the length of analyte zones whether neutral or charged, and in the presence or absence of electroosmotic flow, is solely dependent on the retention factor and the length of the initial zone. These are given by eqs 8 and 15 for neutral and charged analytes, respectively. The sweeping phenomenon is therefore useful for all types of analytes as long as the retention factor is high. Depending on the analytical problem, additives can be added to the sample matrix (organic solvent or nonionic surfactant), pH of sample and separation buffer can be chosen, and a coated capillary can be used to improve the performance of sweeping. Finally, this is another great step in shaping capillary electrophoretic techniques, without alteration of present commercial instrumentation, as a powerful and versatile analytical tool.
1644 Analytical Chemistry, Vol. 71, No. 8, April 15, 1999
ACKNOWLEDGMENT The authors are thankful to Drs. K. Otsuka and N. Matsubara for their support. J.P.Q. is also grateful to the Ministry of Education, Science, Culture, and Sports, Japan (MONBUSHO) and the Japanese Society for the Promotion of Science for supporting his Ph.D. studies. This work was supported in part by a grant-in-aid for Scientific Research (09304071) from the MONBUSHO.
Received for review September 29, 1998. Accepted January 29, 1999.
AC9810866
