Anal. Chem. 1985, 57, 2745-2748 (2)
Burguera, M.; Bogdanskl, S. L.;Townshend, A.
CRC Crit. Rev. Anal. Chem. 1880, 1 1 , 185-246. (3) . Belcher, R.: Boadanskl, S. L.; Knowles, D. J. Townshend, A. Ana/. Chim. Acta 187% 77, 53-63. (4) Bogdanskl, S. L.; Townshend, A,; Blanco, P. T. Anal. Chim. Acta 1881, 131, 297-301.
2745
( 5 ) Koumltzls, A. Anal. Chim. Acta 1877, 88, 303-311. ( 6 ) Belcher, R.; Bogdanskl, S. L.; Knowles, D. J.; Townshend, A. Anal. Chlm. Acta 1975, 7 9 , 292-295.
RECEIVED for review June 10,1985. Accepted July 22,1985.
Electrochemlcal High-Performance Liquid Chromatographic Detection in Aqueous and Nonaqueous Eluents without Supporting Electrolyte Gary
W.Schieffer
Analytical Chemistry Department, Norwich Eaton Pharmaceuticals, Inc.,’ Norwich, New York 13815 Because of its selectivity and sensitivity, electrochemical detection has been increasingly used in high-performance liquid chromatography (HPLC). For reasons of conductivity, the great majority of work has been restricted to reversedphase or ion-exchange systems employing aqueous/organic modifier eluents with added supporting electrolyte salt. However, for reasons of simplicity, ease of extracting the analyte peak by solvent evaporation, or employment of the selectivity advantages of normal-phase separations, the ability to use less-conducting mobile phases (especially those without added supporting electrolyte) might be desirable. Using a conventional small-volume wall jet cell, Gunasingham and Fleet (I) were able to detect nanogram quantities of phenols using a normal-phase eluent of 85:15 hexane/ ethanol containing a t least 0.01 M tetrabutylammonium fluoroborate as supporting electrolyte. Using a large-volume wall jet cell, Gunasingham, Tay, and Ang (2) were able to use a similar eluent system free of supporting electrolyte by making the supporting electrolyte addition postcolumn. However, nearly a 10-fold loss in sensitivity resulted. Recently Evans and Kaaret (3) described a stationary electrochemical cell in which a screen electrode was impressed on a Nafion 117 ion-exchange membrane. The electrode side faced a supporting electrolyte-free aqueous or nonaqueous solvent containing the electroactive species, while the back side faced an aqueous electrolyte solution containing counter and reference electrodes. The transport of charge by ionic conduction occurred through the “solid polymer electrolyte” membrane, allowing electrochemical experiments to be performed without adding supporting electrolyte to the sample solution. The present work explores the feasibility of employing a similar electrochemical cell with a solid polymer electrolyte as an HPLC detector. The main advantage of such a detector might be the ability to operate in reversed-phase and normal-phase eluents without addition of supporting electrolyte, postcolumn addition of reagents, or a significant decrease in signal-to-noise ratio (increase in detection limit). EXPERIMENTAL SECTION Electrochemical Cell. The cell, used initially without modification, has been described previously (Figure 2 of ref 4). It consisted basically of crushed reticulated vitreous carbon (RVC) (Fluorocarbon Co., Anaheim, CA) packed in a 2.2-cm by 2.8mm4.d. tube made from a cation-exchange membrane (0.25 mm thick, Ndion XR-170, Du Pont, Inc., Wilmington, DE). The tube was placed in a Plexiglas chamber containing 0.1 M potassium chloride and a platinum counter electrode and a silver-silver chloride reference electrode. After the initial experiments, the cell was transformed into an annular porous electrode by centering ‘ A Procter & Gamble Company. 0003-2700/85/0357-2745$0 1.50/0
a 2.2-cm by 1.6-mm-0.d.piece of microbore Teflon tubing in the membrane tube and packing the crushed RVC around it, BS shown in Figure 1. The 0.3-mm4.d. channel of the Teflon tube was allowed to become at least partially plugged with RVC particles so that only a small fraction of sample solution avoided working electrode contact. Electrochemical Instrumentation. An EG&G Princeton Applied Research (PAR, Princeton, NJ) Model 174A polarographic analyzer was used to apply potentials to the crushed RVC cell, The 10-V output of the analyzer was converted to 1 V with a voltage divider and fed to a computing integrator (Model 3357, Hewlett-Packard, Avondale, PA) through an analog-to-digital (A/D) converter (Hewlett-Packard Model 18652A). The output of the computing integrator in pV s was converted to coulombs through knowledge of the analyzer’s current range setting, the exact resistance values for the voltage divider, and a calculated correction factor for imprecise A/D converter components (5). The value of 1.075 X lo-’’ C/(pV s) obtained for a 10-pA full-scale current range can only be considered approximate since the injector loop volume (vide infra) and analyzer current-to-voltage amplifier were not accurately calibrated as in the previous work (5). Peak base lines were drawn manually with a BASIC program. All chromatograms shown were obtained with the potential of the crushed RVC electrode set in the limiting current/charge region as determined from hydrodynamic voltammograms. Chromatograms were retrieved with a Hewlett-Packard Model 7470A plotter. HPLC Equipment. The HPLC was equipped with a dualpiston reciprocating pump (Laboratory Data Control Model 111, Riviera, Beach, FL) without a pulse dampener and a Rheodyne Model 7120 injector (Rheodyne, Inc., Berkeley, CA) with a nominal 100-pL loop. Reagents, Columns, and Eluents. For a-tocopherol (Sigma Chemical Co., St. Louis, MO), a 4.6-mm by 25-cm reversed-phase column Containing 5-pm CIS packing (Zorbax ODs, Du Pont Instruments, Wilmington, DE) was used with a pure methanol eluent both with and without 0.05 M sodium perchlorate supporting electrolyte. The flow rate was 2.0 mL/min. A 4.6-mm by 25-cm normal-phase column containing 5-pm silica packing (Partisil5, Whatman, Inc., Clifton, NJ) was also used for a-tocopherol with a 93:7 hexane/l-propanol eluent flowing at 1.0 mL/min. The eluent contained no supporting electrolyte. For hydroquinone (Eastman Kodak Co., Rochester, NY)f the reversed-phase column was used at 1.0 mL/min with a 30:70 methanol/water eluent both with and without 0.05 M potassium nitrate supporting electrolyte. For propyl p-hydroxybenzoate (Eastman Kodak), the normal-phase column was used with an 85:15 hexane/ethanol eluent at 1.0 mL/min with no supporting electrolyte. Flow injection analysis was also done on this compound using the 30:’70 methanol/water eluent with 0.05 M potassium nitrate. All analytes were dissolved in the appropriate eluent.
RESULTS AND DISCUSSION Reversed Phase. Hydrodynamic voltammograms for the oxidation of 408 ng of a-tocopherol in methanol and of 43 ng 0 1985 American Chemical Society
2746
ANALYTICAL CHEMISTRY, VOL. 57, NO. 13, NOVEMBER 1985
Figure 1. Annular porous working electrode, drawn to scale: (A) cation exchange membrane, (B) crushed RVC, (C) Teflon tube. A complete crushed RVC electrode assembly is shown in Figure 2 of ref 4. 50
50
0.4
E,V
0.2
E,V
Figure 2. Hydrodynamic voltammograms of (A) 408 ng of a-tocopherol in methanol with and without 0.05 M sodium perchlorate and (B) 43 ng of hydroquinone in 30:70methanoVwater with and without 0.05 M potassium nitrate. Triangles denote electrolyte-free eluent. See text for chromatographic parameters. of hydroquinone in 3070 methanol/water are shown in Figure 2. Both sets of voltammograms were obtained with and without 0.05 M sodium perchlorate or potassium nitrate dissolved in the eluent as supporting electrolyte. The a-tocopherol voltammograms were obtained with the original crushed RVC cell while the hydroquinone voltammograms were obtained with the annular cell. For a-tocopherol, ommission of the supporting electrolyte shifted the half-wave potential about 75 mV toward more positive values and decreased the limiting charge by about 20%. Unexpectedly, ommission of the electrolyte shifted the half-wave potential for hydroquinone about 70 mV toward more negative values, although reduction of the limiting charge was nearly the same as that for a-tocopherol, about 23%. Apparently, a mechanism other than iR drop is at least partially responsible for the potential shifts observed when eluents with and without electrolyte are used. When a-tocopherol was initially examined with the original crushed RVC cell, it was found that the low-conducting electrolyte-free eluent resulted in a badly tailed peak, as shown for peak I in Figure 3. Insertion of the Teflon tube to form a porous annular electrode yielded a more symmetrical peak (peak 11) whose width at half-height was only 19.8% greater than the peak obtained with the annular cell in which supporting electrolyte was added to the eluent (peak IV). It appears that the electrode surfaces farthest from the ion-exchange polymer electrolyte wall provide sites for adsorption when situated in a relatively low-conducting, supporting electrolyte free solvent. The fact that this central region may not be very electrochemically active is indicated by the increase in limiting charge in Figure 2 and peak area in Figure 3 (peaks I and I11 and peaks I1 and IV, respectively) when
electrolyte is added to the eluent. The small difference in area between peaks I and I1 and peaks I11 and IV probably reflects either analyte lost through the 0.3-mm-i.d. bore of the Teflon tube used to form the annulus or the electrode surface area lost by the displaced crushed RVC. The difference in area between peaks I and I11 and peaks I1 and IV is caused solely by the presence or absence of supporting electrolyte. The difference in capacity factors between peaks I and I11 and peaks 11 and IV was reproducible and probably related to modification of the column packing surface by the perchlorate anion. The small difference in capacity factors observed between the original cell and the annular cell (peaks I and I1 and peaks I11 and IV, respectively) was apparently caused by a between-day variation in column condition and/or mobile phase composition. For hydroquinone, the peak observed in the electrolyte-free solvent (peak I1 in Figure 4) was symmetrical with an 18.8% increase in width at half-height over that observed for the peak in the electrolyte-containingsolvent (peak I). The former peak picked up low-frequency noise from the undampened pump pulsations, as shown. This phenomenon tended to occur variably and unpredictably with electrolyte-free eluents but could easily be remedied by employing the RC filtering of the polarographic analyzer (not done for the comparison in Figure 4 to avoid a slight instrumental band broadening). Similar noise pickup has occurred with conventional detectors (6); however, no attempt has been made to follow the precautions listed in ref 7 for obtaining low-noise measurements. A linearity plot obtained from five injections of 3.5-90 ng of a-tocopherol in supporting electrolyte-free methanol with the annular crushed RVC cell at 0.5 V yielded a straight line with a correlation coefficient of 0.9997. A negative deviation from linearity occurred for weights greater than 200 ng. A similar, but slightly smaller, deviation from linearity was observed when supporting electrolyte was added to the eluent. A linearity plot obtained from six injections of 2.6-170 ng of hydroquinone in electrolyte-free methanol/water with the annular electrode at 0.5 V yielded a straight line with a correlation coefficient of 0.9999. Linearity at higher sample weights was not examined. For these systems a detection limit of about 1ng with a signal-to-noise ratio of about 3 was observed in eluents both with and without electrolyte. (Although the sensitivity of crushed RVC detectors in terms of pA/nmol is higher than that for amperometric detectors, the noise generated by the large electrode surface area of the crushed RVC is responsible for a higher detection limit (5).) Normal Phase. Initially, a-tocopherol in the 93:7 hexane/ 1-propanol eluent without supporting electrolyte was examined. Although the annular cell held at 0.6 V initially yielded a symmetrical peak, the peak height rapidly decreased with succeeding injections until the peak all but disappeared. This decrease became greater as the period between injections was allowed to increase. Flushing the cell with 2-propanol followed by water, 2-propanol, and normal-phase eluent restored the peak to nearly its original size. However, the peak height again decreased with each succeeding injection. It appears the loss of response may be related to the tendency of the poorly conductive organic solvents to gradually insulate the electrode material near the wall from the ion-exchange membrane, rather than an electrode fouling effect. In any event, no reproducibility was attainable. Fortunately, this problem of poor reproducibility did not appear to occur for propyl p-hydroxybenzoate in an electrolyte-free, normal-phase eluent of slightly reduced hexane content. A hydrodynamic voltammogram of 480-ng injections of this compound in electrolyte-free 85:15 hexane/ethanol is shown in Figure 5 , along with a chromatogram for a 20-ng injection. A hydrodynamic voltammogram of this compound
ANALYTICAL CHEMISTRY, VOL. 57, NO. 13, NOVEMBER 1985
2747
i
1.06 uC
\
54
I
I
I
0
2
I
I
0
I
1
MINUTES
Figure 3. Chromatograms of 175 ng of a-tocopherol in methanol: (I) orlglnal porous electrode with electrolyte-free eluent, ( I I) annular electrode with electrolyte-free eluent, (111) orlglnal electrode with 0.05 M sodium perchlorate In eluent, (IV) annular electrode with electrolyte in the eluent. Numbers denote peak area in pC (current is 4-5 pA full scale). For clarity, only the apex of peak 111, which was similar In shape to peak IV, Is shown. Chromatograms are normalized by height. See text for chromatographic parameters. 49.1 37.6
nn
(
I 0
I 2
I 4
b* I
1
I
I
2
I
I 4
1
42nA
\
I
I 6
MINUTES
6
4
I
t
I
I 6
MINUTES
Flgure 4. Chromatograms of 43 ng of hydroquinone in 30:70 methanol/water with (I)0.05 M potassium nitrate in the eluent and (11) electrolyte-free eluent. Numbers denote peak area in pC (current Is 1.5-2 pA full scale). Chromatograms are normalized by height. See text for chromatographic parameters.
obtained by flow-injection analysis in the 30:70 methanol/ water eluent containing 0.05 M potassium nitrate was similar in shape (including the limiting charge maximum) to that obtained in electrolyte-free hexane-ethanol. However, in the more conductive solvent, the half-wave potential was shifted about 310 mV toward more negative values and the total limiting charge was increased by a factor of 16. It appears that in the low-dielectric organic solvent, only the electrode surfaces adjacent to the ion-exchange membrane wall are electrochemically active, resulting in a loss of coulometric efficiency. Despite the 16-fold loss in signal, the base line noise decreased from about 20 nA in conducting solvents to only 600 pA in the electrolyte-free hexane-ethanol solvent. Thus the detedion limit did not increase in the normal-phase eluent. A linearity plot obtained from five injections of 4-50 ng of propyl p-hydroxybenzoate in electrolyte-free hexane-ethanol a t 1.5 V yielded a straight line with a correlation coefficient of 0.9987 and a slope of 0.043 pC/ng. A similar plot obtained from four injections of 50-480 ng of analyte yielded a straight line with a correlation coefficient of 0.9986. However, the slope
Flgure 5. Hydrodynamic voltammogram of 480 ng of propyl p hydroxybenzoate and chromatogram of 20 ng of propyl p-hydroxybenzoate in electrolyte-free 85: 15 hexane/ethanoL Numbers denote peak area and peak height. See text for chromatographic parameters.
decreased to 0.032 pC/ng. At the higher weights injected, the peak height decreased gradually (not rapidly as for a-tocopherol) with each succeeding injection. For example, the peak area of a 200-ng injection decreased by about 5% after 1 2 injections. This was probably a result of electrode fouling; a similar phenomenon has been reported for the oxidation of phenols in a hexane-ethanol eluent (I) and propyl p hydroxybenzoate in methanol-water eluent (8),both eluents containing electrolyte. After use with methanol for 2 weeks and hexane-based solvents for 1week, the response of the cell suddenly became very noisy and the presence of a hexane layer became apparent in the 0.1 M potassium chloride reference solution. After the cell was disassembled and the formed cation-exchange membrane tube was dried, it was found that the membrane had turned hard, brittle, and milky white. It was no longer sufficiently flexible to seal the working electrode channel to prevent eluent from entering the aqueous reference compartment. In addition, parts of the Plexiglas housing showed traces of having been etched by the organic solvents. Nevertheless, use of an ion-exchange membrane as a solid polymer electrolyte shows considerable promise for designing electrochemical HPLC detectors for use in both normal- and reversed-phase eluents without the necessity of adding supporting electrolyte either directly or postcolumn and without an increase in the limit of detection. More work needs to be done, however, to optimize electrode geometry and to make the cell easier to use with normal-phase eluents. For example, the improved peak shape observed with the annular cell and the loss of coulometric efficiency observed in poorly conducting, electrolyte-free eluents indicate that designs in which the working electrode is constrained close to the ion-exchange membrane might be necessary for optimum response. A design involving componenta machined from more inert polymers and an easily replaceable ion-exchange membrane that tightly seals the aqueous reference solution from the sample solvent also might increase the useful lifetime of the detector. Evan's group is currently undertaking a thorough study of electrode design and fundamental electrochemical characteristics of cells with a solid polymer electro!yte (9). LITERATURE CITED (1) Gunasingham, Harl.; Fleet, Bernard J . Chromafogr. 1983, 261, 43-53. .
(2) Gunasingham, Hari,; Tay, 8. T.; Ang, K. P. Anal. Chem. 1984, 56, 2422-2426. (3) Evans, Dennis H.; Kaaret, Thomas W. "Book of Abstracts"; 186th National Meeting of the American Chemical Society, Washington, DC; 1983; any1 49.
2748
Anal, Chem. 1985, 57, 2748-2751
(4) Schieffer, Gary W. Anal. Chem. 1980, 52, 1994-1998. (5) Schieffer, Gary W. Anal. Chem. 1985, 57, 968-971. (6) Schieffer, Gary W. J . Chromafogr. 1080, 202,405-412. (7) Behner, E. Dale; Hubbard, Richard W. Clln. Chem. (Winston-Salem, N.C.), 1979, 25, 1512-1513. (8) King, William P.; Kissinger, Peter T. Clin. Chem. ( Winston-Salem,
N.C.), 1080, 26, 1484-1491, (9) Evans, Dennis H., University of Wisconsin-Madison, communication, 1985.
personal
RECEIVED for review June 19, 1985. Accepted July 22, 1985.
Applications of a Computerized Flow Programmer for Capillary Column Gas Chromatography Soren Nygren*’
Department of Analytical Chemistry, University of Uppsala, P.O. Box 531, S-751 21 Uppsala, Sweden Svante Anderson
T h e National Swedish Laboratory for Agricultural Chemistry, S-750 07 Uppsala, Sweden It has been shown (1-4) that exponential flow programming in capillary gas chromatography under isothermal conditions gives peak distributions which are very similar to those obtained with temperature programming. The technique might thus be used as a useful alternative to temperature programming. With flow programming the elution of the higher boiling components will be accelerated by the steadily increasing flow rate of the carrier gas. This leads to more evenly spaced peaks in the chromatogram. Experience shows that peak widths are not adversely affected by the high flow rates at the end of the separation. In fact, the peak widths stay almost constant, but the efficiency is somewhat reduced at the end of a run because the flow rate is not optimal in the last part of the flow program. With mass sensitive detectors the high flow rates are beneficial since the signal to noise ratio increases. Flow programming is particularly well suited for short or wide-bore capillary columns, which permit separations to be achieved in a few minutes (5). Since only the pressure drop over the column needs to be reset before the next run, the sample throughput can be made very high. The technique is, however, not well adapted to packed columns due to the high flow resistance of such columns. The range of boiling points which can be covered in a flow programmed run is relatively large but not as broad as with temperature programming. The two techniques can, however, be combined in order to cover a wide range of boiling points. Since separation can be carried out a t a lower temperature with flow programming than with temperature programming, the former technique is well suited for separations of thermolabile compounds and also reduces bleeding of the stationary phase. No peak order reversals will occur as is sometimes the case with temperature programming. The flow function Q, = Q, + Qoekqtr used for exponential programming gives the relationship between the retention flow rate, Q,, and the retention time, t,. The parameters Q, and Qotogether with the programming rate, K,, determine the range and rate of the program. The sum of Q, and Qo is equal to the starting flow rate. The retention volume, VI, obtained by integrating Q, dt between the limits t = 0 and t = t,, is
V , = Q,t,
+ Qo/kq(ekqtr- 1)
(2)
The present paper describes a flexible, microcomputer conPresent address: Pharmacia AB, s-751 82 Uppsala, Sweden. 0003-2700/85/0357-2748$01.50/0
trolled device which, by controlling the gas pressure, regulates the flow of the eluting gas according to the flow function. The effectiveness of this method is demonstrated by reference to a number of typical applications using various conditions and chromatographic equipment. EXPERIMENTAL SECTION Chromatographic Systems. The chromatographs used were a Pye unicam 204 with a splitless column injection system coupled to a FPD detector, a Pye Unicam GCD equipped with a “falling needle” injector for capillary columns coupled to a FID detector, and a Pye Unicam GCV with a splitless column injection system and an ECD detector (“Ni). The samples separated were a standard mixture of organophosphorus pesticides, two different mixtures of polyaromatic hydrocarbons (PAH), and an extract from Baltic herring containing chlorinated hydrocarbons. The columns used were AmAc, 25 m, 0.40 mm i.d., SE-54, 50 m, 0.30 mm i.d., and SE-30, 25 m, 0.30 mm i.d. Flow Programmer. A general view of the flow programmer is shown in Figure 1. The personal computer was an ABC 80, Luxor, Sweden. The flow rate of the eluting gas can be regulated in two ways. The first method uses the calibration curve of the gas pressure vs. the flow rate. A dc motor connected to the pressure regulator (ZX 40XT, ITT, The Netherlands) adjusts the pressure as determined by the flow function and the calibration curve. The second method uses a pressure transducer (LX 1710G, National Semiconductor). The pressure is measured continuously. The signal from the transducer is converted from analog to digital form and is then compared by the computer with the value set by the flow function. When a difference is detected between the actual and the set pressure, a signal is generated. After amplification the signal activates two relays regulating the dc motor connected to the pressure regulator. When the pressure has reached the desired value, the motor is turned off by short circuiting, which ensures an instantaneous stopping of the motor. The computer program is structured in blocks. The flow function parameters and the final retention time, i.e., the time alloted to the separation, can be entered in one block. The computer then calculates the programmed final retention flow rate and the final retention volume and shows the graph of the flow function. In another block of the program, retention times are calculated from the known retention volumes of the analytes using eq 2. The results are presented on the screen as a simplified chromatogram. During the run the flow rate is calculated from eq 1 and displayed continuously vs. time. The actual values of time, pressure, retention volume, and flow rate are also shown on the CRT as well as the graph of the flow rate vs. time calculated from eq 1. RESULTS AND DISCUSSION The gas flow through the column is regulated by the inlet pressure. When the flow rate is expressed as an average flow 0 1985 American Chemical Society