Anal. Chem. 1904, 56,293-297
mixing chamber and the use of a peristaltic pump may be responsible for the difference. But the difference may also be due to the fact that the bolus size, according to our calculations, amounted to more than 80% of the system volume and the equations are not known to be valid under these conditions. Also, the carrier stream differed slightly from the solvent in which the methyl orange was dissolved and this may add a slight deviation. Despite these deviations from ideality, the results indicate close similarity of performance with the ideal bolus injection predictions. Gerhardt and Adams (3) used eq 2 to determine diffusion coefficients for biogenic amine neuro transmitter related compounds. Rather than use a plot of log At vs. log q , they preferred to calibrate the system with a substance whose diffusion coefficient was known. Then, if the flow rates were the same and the peak heights were the same for an unknown, it can easily be shown from eq 2 that 2.788
5=
Dk
(5)
30rDL
0.002 I-5 0.8 4
(3)
where k refers to known and u to unknown. This approach has merit since the value of the diffusion coefficient is sensitive to the values of a and f i n eq 2. The radius, a, in particular, is difficult to determine and thus the problem is eliminated. It is interesting that they found the peak heights had to be reproduced to within 5% to obtain accurate values of the coefficients. This agrees with the fact that the theoretical expressions were derived in approximately that way (I). Gerhardt and Adams, however, did use a peristaltic pump, a colorimetric detector, and coiled tubing. The coiled tubing did have some effect on the At values, but regardless of coiling, similar values of the diffusion coefficients were obtained. Since eq 3 is valid over a wide range of L and q values, it is possible to choose values such that tube coiling is not necessary. Also the use of a syringe pump would most closely approximate laminar flow and a fluorescence detector would view the stream perpendicular to the flow. To see if these effects influenced the determined diffusion coefficient, the apparatus originally described in ref 1 was used to determine the coefficient for DOPAC (dihydroxyphenylacetic acid). Here, the tubing length was approximately 35 cm and the flow rate varied from 0.04 to 0.8 mL/min. First, the At values for fluorescenesodium in water injected into water as the carrier fluid were determined for the different flow rates. A plot of log At vs. log q yielded a linear regression line ( 4 ) of In At = 2.698 - 0.678 In q. Further experiments indicated that there was no significant variation of this expression on fluorescene-sodium concentration when this solute was injected with water as the carrier stream. The experiment was repeated with DOPAC in 0.1 M phosphate buffer, pH 7.4, injected into a carrier stream of 0.1 M phosphate buffer, pH 7.4. The linear regression line is given by In At = 2.566 - 0.671 In q. Note that the slopes of -0.678, and -0.671 agree reasonably well with
293
the predicted value of -0.64. In all of the above measurements, the sensitivity of the detector was adjusted to obtain equal peak heights. The ratio of the corresponding At values at any given value of q would allow the determination of the diffusion equation by eq 3. If the diffusion coefficient of fluorescenesodium is taken to be 3.67 X 10+ cm2/s (5),then the coefficient for DOPAC is determined to be (5.66 f 0.16) X lo4 cmz/s (95% confidence interval), which agrees to within 3.5% of the value obtained by Gerhardt and Adams. Apparently, the nonlaminar and other effects are not too significant in their experimental setup. It should be mentioned that this coefficient for DOPAC was determined from the At values taken at the average q value since the standard error is minimal there (4).
Although no diffusion coefficient dependence on concentration was found for fluorescene-sodium injected in water, a significant effect was found for a fluorescene solution injected into a carrier stream containing fluorescene. For example, At values were recorded for two different situations: one for a solution of 2.81 mg/L fluorescene injected into water and the other of 2.81 mg/L fluorescene injected into a 0.281 mg/L fluorescene carrier stream. The ratio of the At's at the average value of q indicate that the diffusion coefficient has increased by a factor of 1.78 to the value of (6.52 f 0.03) X lo4 cm2/s (95% confidence interval). The medium into which the solute is diffusing thus is important and such effects can easily be measured by flow-injection techniques. We hope this paper provides clarification of the conditions under which eq 1 and 2 should be valid and indicates further how flow injection systems should be designed when molecular parameters are to be determined or what cautions should be exercised when experimental results are to be compared with theoretical predictions. Registry No. DOPAC, 102-32-9;fluorescene-sodium, 1318227-9.
LITERATURE CITED (1) Vandersiice, J. T.; Stewart, K. K.; Rosenfeld, A. G.; Higgs, D.J. Talanta 1981, 28, 11-18. (2) Alexander, P. W.; Thalib, A. Anal. Chem. 1983. 55, 497-501. (3) Gerhardt, G.; Adams, R. N. Anal. Chem. 1982, 5 4 , 2618-2620. (4) Zar, J. H. "Biostatistical Analysis"; Prentice-Hall: New York, 1974; Chapter 16. (5) Hodges, K. C.; LaMer, V. K. J . Am. Chem. SOC.1948, 70, 722-726.
'
Present address: University of Maryland Medical School, Baltimore, MD 21201.
Joseph T. Vanderslice* Gary R. Beecher A. Gregory Rosenfeld' Nutrient Composition Laboratory Beltsville Human Nutrition Research Center U.S. Department of Agriculture Beltsville, Maryland 20705
RECEIVED for review, August 9, 1983. Accepted October 21, 1983.
Separation of Complex Mixtures by Parallel Development Thin-Layer Chromatography Sir: Significant improvements in thin-layer chromatography (TLC) have occurred over the past 10 years. These improvements have affected the quality of commercially available TLC plates, the sensitivity, precision, and accuracy obtainable with TLC scanners, the convenience and precision of TLC
spotting devices, and the design of TLC development chambers. A major advantage of TLC as compared to HPLC is that a large number of samples can be run on a single plate which results in a very short analysis time per sample. A significant
0003-2700/84/0356-0293$01.50/00 1984 Amerlcan Chemical Society
294
ANALYTICAL CHEMISTRY, VOL. 56, NO. 2, FEBRUARY 1984
disadvantage is that the separating power of a TLC plate is in general less than that of a HPLC column. Samples containing more than about eight components are rarely separated by TLC without resorting to multiple development techniques which are time consuming. An alternative approach is presented below whereby complex mixtures are separated in parallel developments that proceed simultaneously, resulting in a relatively short time for separation. The computational methods used for determining the experimental conditions for the parallel developments can be extended to predict optimum conditions for two-dimensional TLC.
EXPERIMENTAL SECTION Steroids were purchased from the Sigma Chemical Co. (St. Louis, MO). Solvents used were ethyl acetate, 1,1,2-trichlorotrifluoroethane, tetrahydrofuran, and toluene obtained from the Aldrich Chemical Co. (Milwaukee, WI). A Camag Linear Chamber was used for chromatography. The chamber was used either in the conventional mode or in a modified continuous development mode as described elsewhere ( I ) . Analtech plates, No. 46011, were used for the parallel development TLC. Whatman plates, No. 4807-700, were used for the two-dimensional TLC. All plates were maintained at a 60% relative humidity, by storing in a desiccator over 39% (w/w) sulfuric acid solution, until immediately before use. The steroids were visualized as described elsewhere ( I ) . RESULTS AND DISCUSSION There are several approaches that can be used for separating a complex mixture by TLC. The classic approach is to use two-dimensional TLC whereby a mixture is separated in two sequential developments, each performed with a different solvent system and with the plate rotated by 90' between developments. Another approach is to perform repetitive developments in the same direction by using a solvent of low strength (2-4). Spots are converted into narrow bands allowing resolution between adjacent compounds even when separation distances are small. A third approach is to use continuous development whereby solvent is allowed to evaporate off the end of the plate (5). Spots continue to migrate at rates that are R, dependent for as long as the solvent evaporates. Judicious choice of solvent and operating conditions is necessary for all of the above techniques. Such choice is facilitated by describing the position of all components of a mixture as a function of time, plate length and solvent composition as is discussed below. It is illustrated both for a technique for which we suggest the name parallel development thin-layer chromatography and for two-dimensional TLC. In the former technique, solvent systems are chosen in such a way that all components of a mixture are adequately separated on a small number of plates that are simultaneously developed. The mixture is divided into sets of compounds such that each set is completely separated on one of the plates and that each component of the mixture is represented in at least one of the sets. The components of the mixture that are not fully separated on a particular plate will cluster either near the origin or near the solvent front. The selection of sets is performed by constructing plots of distance migrated, for a specific solvent path length, 1, in a given time, tl, vs. mole fraction. Such plots require the calculation of distance migrated as a function of time, solvent path length, and solvent mole fraction. The following approach is used: Firstly the plate length, I, and analysis time, tl, are chosen. The selection of these parameters is discussed further on in this paper. The specified analysis time, ti, is tl = tl + t , (1) where t , is the time during which the solvent front is ascending
the TLC plate and t zis the time for continuous development. MD,the distance migrated by each spot during the specified time is
MD = d l
+ dz
(2) where dl is the distance migrated during time t l and dz is the distance migrated during time tz. The distance dl is
dl = Rf(1- X ) (3) where x is included to allow for the distance between the solvent origin and the spot origin. The value of Rf can usually be predicted as a function of mobile phase composition when the latter consists of a binary mixture of a strong and a weak solvent. For such a binary the relationship between capacity factor, k, and the mole fraction, X,,of the polar component can be expressed as I n k = a In X,+ b (4) This relationship was originally reported by Soczewinski (6) in a slightly different form. This relationship has already been used in solvent and in time optimization studies in TLC (7-9). There is always the possibility of solvent demixing and the formation of secondary solvent fronts when two solvents of significantlydifferent polarity are used as a mixture (IO). Such secondary fronts would be expected to destroy the linearity described in eq 4. This was not encountered in the work described below either due to the fortuitous choice of solvent systems or due to the fact that a secondary front was present but that all spots were behind this front. The relationship between retardation factor, Rf, and capacity factor is Rf = 1/(1 k) (5)
+
Thus R, can be calculated at any mole fraction, X,,of the binary once the constants a and b are established experimentally. During the continuous development, the rate of migration of the solvent is K/21, where K is the velocity constant in mm2/s. From this it follows that the rate of migration of any spot will be R f ( ~ / 2 1 )Thus . the distance dz is
d2 = R f ( ~ / 2 1 ) t ,
(6)
tl, the time for the solvent front to traverse the plate length 1 is
ti =
J2/K
(7)
Rearranging eq 1 it follows that
t, =
ti
- L2/K
(8)
and (9)
Rf can be expressed as a function of capacity factor and mole fraction (eq 5 and 4). From this and eq 2, it follows that
For the sake of brevity, several steps in the derivation of eq 10 are not included but will be described elsewhere. Thus M Dis a function of X,,1, and ti. K is a function of X , as discussed elsewhere ( I ) . A plot of MDvs. X,can be constructed if the values of 1 and t, are specified. Such a plot is shown in Figure 1for the 15 steroids listed in Table I, chromatographed on silica gel using ethyl acetate/l,l,2-trichlorotrifluoroethane as solvent. It is suggested that the value
ANALYTICAL CHEMISTRY, VOL. 56, NO. 2, FEBRUARY 1984
295
0 0
D 0 P
7
Flgure 1. Plots of computed spot migration vs. mole fraction for 15
steroids. Plate length is 100 mm and time is 750 s.
-+-e
cb.45
0.47
0.49
MOLE
OI.53
01.51
FRACTION
OI.55
0
Figure 3. Plots of computed spot migration vs. mole fraction for 15
3
steroids as in Figure 1 but with expanded axes.
0
0 0
0
0 0
z i
0
5 I . ' %
0:17
O'.l9
MOLE
0'.21
1
0'.23
0.26
Figure 2. Plots of computed spot migration vs. mole fraction for 15 steroids as in Figure 1 but with expanded axes.
Table I. Fifteen Steroids Grouped in Sets for Separation by Parallel Development TLC low R f steroids high Rf steroids mestranol cholesterol ergosterol estrone androstanedione progestevone
0
FRACTION
ethisterone androstanolone acetoxyprogesterone epiandrosterone methandriol androst enediol testosterone cortisone digoxin
of 1 be specified as 10 cm for the MD vs. X,plot. The value of tl is chosen such that the compound of highest R, will migrate to the end of the plate a t the highest usable mole fraction. For the TLC system discussed here, this corresponds to pure ethyl acetate, Le., X, = 1.0. Inspection of Figure 1 shows that there are two sets of compounds. The two sets are more clearly shown in Figures 2 and 3 which have expanded axes. Figure 2 shows seven high Rf compounds m d Figure 3 shows ten low R, compounds that can be separated in the respective mole fraction ranges. This allows some flexibility in the assignment of compounds into sets. For the separation that is reported here, six compounds were assigned to the high R, set and nine compounds to the low Rf set. In principle, optimum conditions for the separation of each set of compounds could be obtained by systematically varying 1 and tl in order to obtain a series of MD vs. X , plots. Careful inspection of these plots would identify the optimum
Figure 4. Computer-simulated and experimental chromatograms for the separation of high R, steroids.
conditions of tl,1, and X,for separating each set of compounds. However it is more efficient to treat the separation of each set of compounds as a time optimization problem whereby the conditions that yield a minimum analysis time can be calculated. Thus the utility of a diagram such as Figure 1 is for sorting compounds into sets, rather than for defining optimum chromatographic conditions. The method of selecting optimum conditions for the TLC separation of a mixture of compounds in a minimum time has recently been described (9). Familiarity with the method is assumed in the ensuing discussion. It is necessary to select a value of S D , the center-to-center spot separation for the most difficult to separate pair of compounds in each set. A S D value of 5 mm to 10 mm will usually result in adequate resolution depending on the type of TLC plate used. The two sets of steroids are listed as high R, and low R, compounds in Table I. The conditions that yield a spot separation of at least 5 mm in the minimum analysis time for the high R, set were calculated as X,= 0.22, plate length = 80.73 mm, development time = 55.8 min; for the low R, set the conditions are X , = 0.48, plate length = 82.6 mm, development time = 45.4 min. The chromatograms obtained are shown in Figures 4 and 5 together with the corresponding computer simulated chromatograms. The compounds with migration distances greater than the plate length are not shown. These compounds are compressed into the solvent front. Thus a t least 5 mm must be left between the solvent front and the MDof the highest Rf member of the low R, set. The agreement between the simulated and actual chromatograms is in general good. A sixteenth steroid could have been
296
66
ANALYTICAL CHEMISTRY, VOL. 56, NO. 2, FEBRUARY 1984
0
I
*
I
I
* * I
I
I
I
* I
*
Figure 5. Computer-simulated and experimental chromatograms for the separation of low I?, steroids. included in the mixture but was not shown as its MD was 6 mm greater than predicted even though it was still fully separated from the other components. The most probable cause for this error is in the measurement of the slope and intercept used in eq 4. It should be emphasized that with suitable apparatus, as many as 35 samples on each of two 20 cm long TLC plates could be simultaneously analyzed with this technique. Thus the development time per sample would be in the order of about 2 min. The total analysis time (spotting, development, and quantitation) would be about 5 min per sample. The method is obviously not limited to a single solvent system. For complex mixtures, it may be necessary to optimize a different binary solvent system for each set of compounds. The method is also not limited to using two plates. A larger number of plates would be necessary for a more complex mixture. A larger number of plates would also separate the same mixture in a shorter period of time. For the particular mixture discussed here the analysis time could be reduced from 55.8 min to 45.2 min by dividing the mixture into three sets of compounds. In this particular case the reduction in analysis time would not be worth the inconvenience of running a third TLC plate. The computational procedures that are described above are not limited to parallel developments. The capability of predicting spot position as a function of plate length, solvent composition, and development time should be useful in optimizing other TLC methods such as two-dimensional TLC. Solvents for this technique are usually selected on a trial and error basis, relying on general chromatographic criteria. This is satisfactory for simple mixtures but can be extremely time consuming for complex mixtures. For such mixtures, each of the two solvent systems, if used alone, would generally be incapable of separating several pairs of compounds. Thus, the choice of solvent is dependent on its ability to separate those solute pairs that are not separated by the solvent used for the second development. For the complete separation of a complex mixture it may be necessary to investigate several solvent systems. It could be a tedious task to consider all the combinations of candidate solvents for two-dimensional development. Fortunately the selection and optimization of binary solvents for two-dimensional development can be performed by an extension of the computational techniques discussed earlier in this paper. The position of any spot can be defined in terms of an x and y coordinate, where each coordinate corresponds to the MD value in each of the solvents used for the two-dimensional development. Then the separation distance for each pair of compounds in the two-dimensional development, S,,yPq, is
sx,p= [ ( s x p q ) 2 + ( S y P q ) 2 ] 1 / 2
(11)
where pq designates the pair of compounds being separated
B O 0
0
0
00 0
0
0 0 0
0
0
Figure 6. Computer-simulated and experimental chromatograms for two-dimensional thin-layer chromatography. and S , P q and S y p q would be the corresponding SDvalues for the pair of compounds p and q in each of the two developments if they were performed separately. The optimization of this method will be discussed elsewhere. It should be noted that a method for optimizing two-dimensional TLC has recently been published (11). This differs from the method described here in that neither continuous development nor optimized binary solvents were used. The authors do however propose using binaries in future work. To illustrate the validity of our method we present the two-dimensional separation of a mixture of 13 steroids on a silica gel plate with a predicted minimum spot separation of 5 mm. The solvents used were tetrahydrofuran/toluene and ethyl acetate/l,l,2-trichlorotrifluoroethane.The predicted separation is shown in Figure 6A and the separation actually found in Figure 6B. In general there is a very good agreement between the two figures. Both runs should be performed at the same humidity. A better separation than that shown could have been obtained if solvents were chosen such that the order of MD’S would be very different for each of the two binary solvent systems. This would have had the effect of spreading the spots over a wider area of the TLC plate than that shown in Figure 6. The separation shown was performed in a modified Camag Linear Chamber. The design of this apparatus results in the constraint that only one of the two developments could be performed in the continuous mode. To somewhat compensate
Anal. Chem. 1984. 56,297-298
for this limitation, a strong solvent (tetrahydrofuran/toluene) in which solutes would have a reasonably high mobility, was used for the noncontinuous development. The Regis SB/CD Chamber should also be useful for two-dimensional TLC. While the length of the plate cannot be fully optimized, continuous development can be performed in both directions. The use of the SB/CD chamber for optimized continuous development TLC has recently been described (8). The results presented in this paper lead us to conclude that careful optimization of operating conditions for continuous development TLC using binary solvents should allow the routine analysis of moderately complex mixtures by TLC.
LITERATURE CITED Nurok, D.; Becker, R. M.; Sassic, K. A. Anal. Chem. 1982, 5 4 , 1955-1959. Thoma, J. A. Anal. Chem. 1963, 35, 214-224. Jupille, T. H.; Perry, J. A. Science 1978, 194, 288-293. Nurok, D.; Zlatkis, A. Carbohydr. Res. 1980, 8 1 , 167-172. Mottier, M.; Potterat, M. Anal. Chim. Acta 1955, 13, 46-56. Soczewinski, E.; Golkiewicz, E.; Szumila, H. J . Chromafogr. 1969, 45, 1-13.
297
(7) Nurok, D.; Richard, M. J. Anal. Chem. 1981, 5 3 , 563-564. (8) Tecklenburg, R. E., Jr.; Becker, R. M.; Johnson, E. K.; Nurok, D. Anal. Chem. 1983, 55, 2196-2199. (9) Tecklenburg, R. E., Jr.; Maidak, B. L.; Nurok, D.HRC CC, J . High Res. Chromatogr Chromatogr . Commun. 1983: 6 , 627-628. (10) Touchstone, J. C.; Dobbins, M. F. Practice of Thin Layer Chromatography"; Wiley: New York, 1978. (1 1) Gonnord, M. F.; Levi, F. J.; Guiochon, G. J . Chromatog. 1983, 264, 1-6.
.
'
Present address: Department of Medical Genetics, Indiana University School of Medicine, Indianapolis, I N 46223.
David Nurok* Ronald E. Tecklenburg, Jr. Bonnie L. Maidak' Department of Chemistry Indiana University-Purdue University a t Indianapolis P.O. Box 647 Indianapolis, Indiana 46223 RECEIVED for review August 12, 1983. Accepted November 4, 1983. This work was supported by grants from the Dow Chemical Co. and the Society for Analytical Chemists of Pittsburgh.
Chemical Pretreatment of Silver Wire Containing Copper for Preparation of Silver/Silver Sulfide Ion Selective Electrode Sir: The silver/silver sulfide ion selective electrode (ISE) has been produced in a variety of forms (1-14). While the membrane type (2-12) reportedly exhibits superior performance characteristics in comparison to the type derived from solid silver (3), the latter offers decided advantages in terms of cost and convenience. We wish to report that a microscale silver/silver sulfide ISE can be readily prepared from silver wire (0.7 mm diameter) which contains 2.5-6% Cu (w/w) by exposing the wire to a deaerated aqueous solution of sodium sulfide, ascorbic acid, and ethylenediaminetetraaceticacid (SAOB-11) (15) for 2 days. Although this procedure is similar to one reported in the manufacturers' literature (15, 16), it is significant that a successful ISE could not be prepared by this technique if pure silver (99.99%) was used or if the SAOB-I1 solution was not deaerated. Finally, the performance of electrodes obtained in this manner was basically indistinguishable from that of a commercially available electrode of the membrane type (Orion Research, Inc., Model 94-16). EXPERIMENTAL SECTION Reagents. All chemicals were of analytical grade and were used without further purification. Calibration solutions were prepared from silver nitrate or sodium sulfide nonahydrate. The standard sulfide solutions were prepared in either SAOB-I1 (15) or 0.1 M NaOH which contained ascorbic acid (20 ppt). The concentration of the sulfide standards was confirmed by potentiometric titration with 0.1 M PbNO, using a commercially available ISE (see above) as the sensor in combination with a double junction calomel reference electrode (Orion Research, Inc., Model 90-02). The outer filling solution consisted of an aqueous solution of KNO, (10% w/w). Preparation of the Ag/Ag,S wire electrode. A silver wire (0.7 mm diameter) was sealed within a glass tube with epoxy cement (Elmers) so that only the end of the wire was exposed. After the cement had dried, the end of the wire was polished with fine grade emery paper. Then, the electrode was immersed in
0.1 M Na2S in SAOB-I1for 2 days. A thin uniform film of Ag2S gradually formed during this period. The nature of the surface was confirmed by scanning electron microscopy and energy dispersive X-ray fluorescence. Finally, the electrodes were rinsed with distilled deionized water and stored in saturated Ag2Ssolution prior to use. If pure silver wire was employed as the substrate or the sulfide solution was not deaerated, no Ag2Slayer formed even after 6 days of exposure. Procedure. All potentiometric studies were carried out with a digital voltmeter (Orion Research, Inc., Model 601A or Fluke Model SOOOA) in a double-walled thermostated cell which was maintained at 25.0 f 0.1 "C. Connections to the electrodes were made via an electrode switch (Orion Research, Inc., Model 605). The double junction reference electrode mentioned above was employed. All pH measurements were made with a common glass electrode (Sargent Model 3-30050-15C). The measurements of response to Ag+ were made upon serial dilutions of a standard 0.1 M AgNO, solution with 0.1 M NaN0,. Similarly, sulfide ion standards were prepared by serial dilution of the 0.1 M stock Na2Ssolution with 1M NaOH solution which contained ascorbic acid at the 20 ppt level. Potentiometric titrations were carried out in the conventional manner. The titrant was delivered in 0.1-mL increments with a 2.0-mL syringe buret (Gilmont).
RESULTS AND DISCUSSION Optimal electrode characteristics in terms of reproducibility and the extent of the linear relationship between potential and concentration were obtained when the electrode was exposed to the deaerated SAOB-I1 solution for no more nor less than 2 days. The requirement for adulterated silver wire suggests that the formation of the Ag2S layer is a t least dramatically facilitated by some anodic corrosion process ( I 7). Once prepared, the electrodes can be stored for long periods under saturated silver sulfide solution or briefly ( 1 week) in air with no noticeable effect. The equilibrium potentials observed during the successive dilution of standard solutions as described above are depicted in Figure 1. Linear response extends over a t least 4 decades
0003-2700/84/0356-0297$01.50/00 1984 American Chemical Society
