Supercritical Fluid Chromatography with Electrochemical Detection of

---

Anal. Chem. 1996, 68, 3121-3127

Supercritical Fluid Chromatography with Electrochemical Detection of Phenols and Polyaromatic Hydrocarbons Shawn F. Dressman, Anna M. Simeone, and Adrian C. Michael*

Department of Chemistry, University of Pittsburgh, Pittsburgh, Pennsylvania 15260

A self-contained electrochemical cell consisting of a working and a counter electrode coated with a thin poly(ethylene oxide) film containing lithium triflate has been evalulated as an on-line detector for supercritical fluid chromatography. Electrochemical detection of 11 priority phenols and 13 polyaromatic hydrocarbons separated on a packed diol column with neat and methanol-modified CO2 mobile phases is described. The detector is operated voltammetrically, which permits the simultaneous detection of both oxidizable and reducible components of individual mixtures. Quantitative aspects of the detector performace have been evaluated and compared to those of a downstream flame ionization detector. The electrochemical detector provides low-nanogram detection limits and responds linearly over two decades of injected quantities. An advantage of voltammetric detection is the selectivity derived from the range of applied potentials used. Overlapping elution profiles of chloro-, methyl-, and nitro-substituted phenols can be resolved voltammetrically by capitalizing on differences between their half-wave potentials. The detector can be used for a period of at least 3 days without refurbishing. The electrochemical detector is compatible with CO2 containing 1% (v/v) methanol, which can be used to decrease the total analysis time for the phenol mixture from 50 min in unmodified CO2 to 20 min. Supercritical CO2 receives a great deal of attention due in part to its ability to extract a variety of compounds from, for example, environmental and industrial matrices.1-5 In addition, CO2 is safe, easy, and economical to use. Hawthorne has pointed out that SFE is an especially powerful analytical tool when it is directly coupled to a chromatographic analysis, mainly because the entire extract can be introduced to the chromatographic column.6 Although the on-line coupling of SFE to GC and HPLC have been reported,6-12 the coupling of SFE to SFC has several distinct (1) Laitinen, A.; Michaux, A; Aaltonen, O. Environ. Technol. 1994, 15, 715727. (2) Hawthorne, S. B.; Miller, D. J.; Burford, M. D.; Langenfeld, J. J.; EckertTilotta, S.; Louie, P. K. J. Chromatogr. 1993, 642, 301-317. (3) Janda, V.; Bartle, K. D.; Clifford, A. A. J. Chromatogr. 1993, 642, 283-299. (4) Hedrick, J. L.; Mulcahey, L. J.; Taylor, L. T. Mikrochim. Acta 1992, 108, 115-132. (5) Hawthorne, S. B. Anal. Chem. 1990, 62, 633A-642A. (6) Burford, M. D.; Hawthorne, S. B.; Miller, D. J. J. Chromatogr. 1994, 685, 79-94. (7) Greibrokk, T. J. Chromatogr. 1995, 703, 523-536. (8) Sandra, P.; Kot, A.; Medvedovici, A.; David, F. J. Chromatogr. 1995, 703, 467-478. S0003-2700(95)01188-7 CCC: $12.00

© 1996 American Chemical Society

advantages. For example, since the extraction solvent and mobile phase are the same, every component of the extract is amenable to SFC, so the extraction as well as the separation is carried out with a single pump. In addition, the coupling of SFE to SFC can be done in a highly flexible manner, since static and dynamic extraction modes, modified and unmodified mobile phases, and packed and capillary columns can be used as needed. To capitalize on the synergism of SFE and SFC, new detectors for SFC are required because existing detectors either lack the necessary sensitivity or are incompatible with the wide range of valuable separation conditions in SFC. The focus of the work described in this paper, therefore, is an on-line electrochemical detector for SFC. Recently, we demonstrated that the electrochemical detector, when coupled to packed column SFC, provides nanogram detection limits and is compatible with both neat and modified CO2-based mobile phases.13 The issues addressed in this paper concern the scope of SFC coupled to electrochemical detection, the long-term stability of the detector response, and the selectivity derived from voltammetric detection rather than constant-potential amperometric detection. Two classes of compounds that, by virtue of their status as priority pollutants and their compatibility with CO2-based fluids, are frequent targets of SFE and SFC are phenols and polyaromatic hydrocarbons (PAHs).14-18 These compounds must be routinely monitored in the environment because of their toxicity and their ubiquitous presence in polluted soils and wastewater.19-23 Although most phenols and PAHs are electroactive, the passivation of electrode surfaces by such compounds presents a formidable (9) Johansen, H. R.; Becher, G.; Greibrokk, T. Anal. Chem. 1994, 66, 40684073. (10) Raymer, J. H.; Velez, G. R. J. Chromatogr. Sci. 1991, 29, 467-475. (11) King, J. W. J. Chromatogr. Sci. 1990, 28, 9-14. (12) Andersen, M. R.; Swanson, J. T.; Porter, N. L.; Richter, B. E. J. Chromatogr. Sci. 1989, 27, 371-377. (13) Dressman, S. F.; Michael, A. C. Anal. Chem. 1995, 67, 1339-1345. (14) Barnabas, I. J.; Dean, J. R.; Tomlinson, W. R.; Owen, S. P. Anal. Chem. 1995, 67, 2064-2069. (15) Honer, A; Arnold, M.; Husers, N.; Kleibohmer, W. J. Chromatogr. 1995, 710, 129-137. (16) Hawthorne, S. B.; Miller, D. J. Anal. Chem. 1994, 66, 4005-4012. (17) Hills, J. W.; Hill, H. H.; Hansen, D. R.; Metcalf, S. G. J. Chromatogr. 1994, 679, 319-328. (18) Berger, T. A.; Deye, J. F. J. Chromatogr. Sci. 1991, 29, 54-59. (19) Achilli, G.; Cellerino, G. P.; d’Eril, G. M.; Bird, S. J. Chromatogr. 1995, 697, 357-362. (20) Lamprecht, G.; Huber, J. F. K. J. Chromatogr. 1994, 667, 47-57. (21) Galceran, M. T.; Jauregui, O. Anal. Chim. Acta 1995, 304, 75-84. (22) Di Corcia, A.; Marchese, S.; Samperi, R. J. Chromatogr. 1993, 642, 175184. (23) Brudzewski, J.; Shusterman, D.; Becker, C.; Borak, J.; Cannella, J.; Goldstein, B.; Hall, A.; Jackson, R. J.; Rodnick, J.; Wheater, R.; Wummer, B. Am. Fam. Physician 1993, 47, 623-628.

Analytical Chemistry, Vol. 68, No. 18, September 15, 1996 3121

challenge to stable electrochemical detection.24-26 This paper demonstrates stable electrochemical detection of 11 priority phenols and 13 priority PAHs separated by packed-column SFC. EXPERIMENTAL SECTION Reagents. SFC-grade carbon dioxide with a He headspace (Scott Specialty Gases, Plumsteadville, PA), H2 and compressed air (Valley Welding, Malvern, PA), and methanol (Mallinckrodt, Paris, KY) were used as received. Phenols were obtained in premixed form and as individual solutions from Supelco (Bellefonte, PA). The following phenols were included in the mixture: pentachlorophenol, 2412 mg/L; phenol, 489 mg/L; 2-chlorophenol, 488 mg/L; 2-methyl-4,6-dinitrophenol, 2401 mg/L; 2-nitrophenol, 492 mg/L; 2,4-dichlorophenol, 492 mg/L; 2,4-dimethylphenol, 495 mg/L; 2,4-dinitrophenol, 1436 mg/L; 2,4,6-trichlorophenol, 1471 mg/L; 4-chloro-3-methylphenol, 2445 mg/L; and 4-nitrophenol, 2409 mg/L. PAHs were also obtained in premixed form and as individual solutions from Supelco. The following PAHs, each at a concentration of 500 mg/L in methylene chloride, were included in the mixture: acenaphthylene, anthracene, benz[a]anthracene, benzo[a]pyrene, benzo[b]fluoranthene, benzo[ghi]perylene, benzo[k]fluoranthene, chrysene, dibenzo[a,h]anthracene, fluorene, indeno[1,2,3-cd]pyrene, phenanthrene, and pyrene. Poly(ethylene oxide) (PEO; Polysciences, Warrington, PA) and lithium triflate (Aldrich, Milwaukee, WI) were used as received. Tris(2,2′bipyridyl)ruthenium(II) hexafluorophosphate [Ru(bpy)3(PF6)2] was prepared by metathesis of Ru(bpy)3Cl2 (Aldrich) and ammonium hexafluorophosphate in water. All phenol solutions were made in methanol and injected using a 1 µL sample loop. PAH solutions were made in dichloromethane (Mallinckrodt). Electrochemical Detector. The procedure for making the electrochemical detector has been described before.13 Briefly, the detector consists of a platinum disk (radius 5 µm) working electrode sealed inside a platinum tube (o.d. 600 µm, i.d. 500 µm) counter electrode with epoxy. The detector was polished using a series of polishing grades beginning with fine sandpaper and finishing with 0.3 µm particle-sized alumina. Prior to placement in the SFC oven, the detector was polished with 0.3 µm alumina (Buehler, Lake Bluff, IL), rinsed and sonicated in deionized water, and dried in air. The detector (i.e., both electrodes) was dipcoated in a solution of either PEO (90 mg/mL), lithium triflate (20 mg/mL), and Ru(bpy)3(PF6)2 (4.3 mg/mL) (Ru/PEO-Li electrodes) or just PEO and lithium triflate (PEO-Li electrodes) in 9:1 (v/v) MeCN/MeOH and dried in air for 60 min. (In the course of the phenol/PAH experiments, we learned that PEOLi electrodes lacking the ruthenium complex are sufficiently conductive to use in unmodified CO2 because temperatures close to the Tm of PEO (60-70 °C) are being used. The data reported herein were obtained with Ru/PEO-Li and PEO-Li electrodes.) The detector was then inserted into a Swagelok low dead volume union tee and was mounted in the oven such that column effluent was directed toward the detector tip and then out the side arm of the union tee to a downstream FID. Supercritical Fluid Chromatography. For all experiments, an MPS 225 system (Suprex, Pittsburgh, PA) was used. The (24) Dubois, J. E.; Lacaze, P. C.; Pham, M. C. J. Electroanal. Chem. 1981, 117, 233-241. (25) Bruno, F.; Pham, M. C.; Dubois, J. E. Electrochim. Acta 1977, 22, 451457. (26) Bejerano, T.; Forgacs, C.; Gileadi, E. J. Electroanal. Chem. 1970, 27, 6979.

3122

Analytical Chemistry, Vol. 68, No. 18, September 15, 1996

system consists of an automated syringe pump, an automated injector with a 1 µL loop, an oven, and a flame ionization detector. The oven temperature was maintained at 60 °C throughout these experiments. For separations with 1% (v/v) methanol-modified mobile phase, 2 mL of methanol was pipetted directly into the syringe pump prior to filling the pump to a volume of 200 mL with CO2. Mixing was achieved by moving the piston up and down 3-4 times after filling with CO2. Voltammetric and Amperometric Detection. The working and counter electrodes of the electrochemical detector were connected to an EI 400 potentiostat (Ensman Instruments, Bloomington, IN) for voltammetric experiments. Electrochemical data were obtained on a PC through software developed in-house. A sweep rate of 25 V/s and a scan frequency ranging from 20 to 60 voltammetric scans per minute were used for experiments involving cyclic voltammetry. A potential range from -1.5 to +2.0 V vs the Pt counter electrode was used for experiments involving the simultaneous oxidation and reduction of phenols. A potential range from 0 to +2 V vs the Pt counter electrode was used for experiments involving only the oxidation of phenols or the oxidation of PAHs. Chromatograms were obtained from voltammetric data by measuring current values from software-selectable potential ranges and plotting the current values as a function of time. PAH chromatograms, for example, were obtained by measuring currents from the 1.8-2.0 V potential range on the positive-going sweep of each voltammogram. Voltammograms were collected at 2 s intervals, such that PAH chromatograms contained 30 data points/min. For amperometric detection, the working electrode was connected to a Keithley 427 amplifier (Keithley Instruments, Cleveland, OH). The working electrode was held at a constant potential of +2 V vs the Pt counter electrode for phenol detection. Current values were collected at 100 ms intervals. Safety Considerations. Mobile phase pressures as high as 350 atm were used in these experiments. To ensure safety during SFC experiments, all equipment including the electrochemical detector were carefully checked for leaks at pressures up to 400 atm prior to these experiments. RESULTS AND DISCUSSION Voltammetric Detection of Phenols in Unmodified CO2. Figure 1 shows chromatograms of a standard phenol mixture obtained with the electrochemical detector (Figure 1A) and with the downstream FID. The identity of each peak and the amount of each phenol in the sample is given in the figure legend. The pressure at the pumphead was maintained at 100 atm for the first 4 min of the separation. The pressure was increased linearly to 350 atm over the next 15 min and was held at 350 atm for the remainder of the separation. The top chromatogram in Figure 1A was obtained by measuring the reduction currents between -0.7 and -1.0 V during the negative-going potential sweep of each voltammogram. The bottom chromatogram in Figure 1A was obtained by measuring the oxidation currents between 1.8 and 2.0 V during the positive-going potential sweep of each voltammogram. Figure 1A shows that all 11 phenols can be voltammetrically detected in unmodified CO2. The elution order is consistent with a previous report by Berger and Deye on SFC of phenols on a diol column.18 All of the phenols generate an oxidation current in the 1.8-2.0 V applied potential range. When reduction currents are measured over a potential range of -0.7 to -1 V, the only

Figure 1. Chromatograms of an undiluted EPA 604 standard phenol mixture obtained with the electrochemical detector (A) and an FID (B). The numbered peaks correspond to the following compounds: (1) 2-nitrophenol, 492 ng; (2) 2-chlorophenol, 488 ng; (3) 2-methyl4,6-dinitrophenol, 2401 ng; (4) 2,4-dinitrophenol, 1436 ng; (5) 2,4dichlorophenol, 492 ng; (6) 2,4,6-trichlorophenol, 1471 ng; (7) 2,4dimethylphenol, 495 ng; (8) phenol, 489 ng; (9) pentachlorophenol, 2412 ng; (10) 4-chloro-3-methylphenol, 2445 ng; and (11) 4-nitrophenol, 2409 ng. The upper chromatogram in (A) was obtained using a potential range from -0.7 to -1.0 V. The lower chromatogram in (A) was obtained using a potential range from 1.8 to 2.0 V.

peaks obtained are those corresponding to the nitrophenols. For three of the four nitrophenols in the mixture (peaks 1, 3, and 4), the absolute value of peak height and the signal-to-noise ratio are noticeably larger when the peaks are obtained using reduction current instead of oxidation current. Lower detection limits for these nitrophenols are therefore obtained by measuring reduction current. For 4-nitrophenol (peak 11), the oxidation and reduction peak heights are similar in magnitude. Figure 1B shows the chromatogram obtained with the FID. The average peak widths measured at half-height and the average peak asymmetry factors measured at 10% of the peak differ by less than 10% between the two detectors. This demonstrates that the polymer film on the electrode does not significantly affect chromatographic peak shapes, which is consistent with our initial studies.13 Two striking differences exist between the electrochemical and FID chromatograms. First, the solvent front obtained with the electrochemical detector is more narrow than the solvent front obtained with the FID. In previous work, we showed that the electrochemical detector produces no solvent front when dichloromethane is the injection solvent.13 In Figure 1, however, the injection solvent is methanol. The cause of the methanol solvent front has not been investigated, but two likely contributors are a change in the conductivity of the polymer film and the direct oxidation of methanol at the working electrode.27 A change in PEO-Li film conductivity by plasticizing vapors was described by Murray.28 The solvent front at the electrochemical

Figure 2. Chromatograms of a diluted EPA 525 standard PAH mixture obtained with the electrochemical detector (A) and an FID (B). The numbered peaks correspond to the following compounds (100 ng each): (1) acenaphthylene, (2) fluorene, (3) phenanthrene, (4) anthracene, (5) pyrene, (6) benzo[a]anthracene, (7) chrysene, (8) benzo[b]fluoranthene, (9) benzo[k]fluoranthene, (10) benzo[a]pyrene, (11) dibenzo[a,h]anthracene, (12) benzo[ghi]perylene, and (13) indeno[1,2,3-cd]pyrene.

detector in Figure 1 nonetheless returns to baseline in about half the time it takes for the FID response to return to baseline. The narrow solvent front at the electrochemical detector allows baseline resolution of early-eluting peaks that overlap with the FID solvent front, such as 2-nitrophenol and 2-chlorophenol (peaks 1 and 2, respectively). The second difference in the chromatograms obtained with the two detectors is the variability in the peak heights of the individual phenols. The FID is a mass-sensitive detector and responds proportionately to the different amounts of the individual phenols in the injected mixture. The electrochemical detector, however, has a specific sensitivity factor for each compound, depending on factors such as the compound’s electrode kinetics, redox potentials, and mass transport properties in the fluid and in the film covering the electrode. This is discussed later in more detail. Voltammetric Detection of Polyaromatic Hydrocarbons Separated by SFC. Figure 2 shows chromatograms obtained with the electrochemical detector (A) and the FID (B) during elution of priority PAHs from a packed diol column with unmodi(27) Encyclopedia of Electrochemistry of the Elements; Bard, A. J., Lund, H., Eds.; Marcel Dekker, Inc.: New York, 1978; Vol. XI, pp 181-340. (28) Reed, R. A.; Geng, L.; Murray, R. W. J. Electroanal. Chem. 1986, 208, 185193.

Analytical Chemistry, Vol. 68, No. 18, September 15, 1996

3123

fied CO2. For the first 2 min of the separation, a constant pressure of 110 atm was maintained. For the next 14 min, the pressure was increased linearly to a final pressure of 250 atm. For the electrochemical data, currents were measured over a potential range from 1.8 to 2.0 V of the positive-going sweep of each voltammogram. The PAHs range in size from three to six fused rings; the order of elution is roughly according to increasing size. Although there are 13 PAHs present in the mixture, only 11 peaks are obtained with each detector. Peak 3 is due to the co-elution of phenanthrene and anthracene and peak 7 is due to the co-elution of benzo[b]fluoranthene and benzo[k]fluoranthene. Each of these was individually injected and voltammetrically detected using the same parameters as in Figure 2. All of the PAHs in the mixture have therefore been detected voltammetrically in unmodified CO2. Two major differences exist between the chromatograms shown in Figures 1 and 2. First, the solvent front obtained with the electrochemical detector is much smaller because dichloromethane was used instead of methanol as the injection solvent. This shows that the size of the solvent front obtained with the electrochemical detector is dependent upon the nature of the solvent. Second, all of the peaks obtained with the electrochemical detector for the PAH injection are similar in area, excluding those which are unresolved. This indicates that the electrochemical sensitivities to PAHs are very similar. Together, Figures 1 and 2 show that voltammetric detection in SFC is applicable to compounds that are frequent targets of SFE and SFC studies. A total of 29 compounds encompassing organometallics (ferrocene), quinones,13 and now phenols and PAHs have been electrochemically detected thus far in CO2-based mobile phases. Voltammetric Detection of Phenols in MeOH-Modified CO2. The remainder of this paper is a detailed study of the voltammetric detection of phenols. SFC of moderately polar compounds such as phenols is usually performed with modified mobile phases, which provide faster elution times than neat CO2.18 It is important, therefore, that electrochemical detection of phenols is possible in modified mobile phases. Figure 3 shows chromatograms obtained voltammetrically during the elution of phenols with 1% (v/v) methanol-modified CO2. The top and bottom chromatograms were obtained by measuring currents of individual voltammograms from -0.9 to -1.2 V of the negative-going sweep, and from 1.8 to 2.0 V of the positive-going sweep, respectively. The addition of 1% methanol to CO2 decreases the retention times for all of the phenols. The retention time of 4-nitrophenol, the last compound to elute in both mobile phases, decreased from 49 min in unmodified CO2 to 19 min in methanol-modified CO2. The same pressure program from Figure 1 with the modified mobile phase did not completely resolve the phenols. Nevertheless, performance of the detector in neat and modified mobile phases can still be compared. The sensitivity of the electrochemical detector is increased in the presence of methanol, consistent with prior observations that plasticizing vapors increase mass transport rates and decrease ohmic resistance in the PEO films.28 The signal-to-noise ratio, however, is slightly better in the neat mobile phase. The signal-to noise ratios of the 4-nitrophenol peaks in Figures 1A and 3, for example, are 193 and 121, respectively. Thus, electrochemical detection of phenols is not significantly compromised by the use of a modified mobile phase, even though the FID is rendered completely unusable under these conditions by the response to the modifier itself. 3124 Analytical Chemistry, Vol. 68, No. 18, September 15, 1996

Figure 3. Chromatograms of undiluted EPA 604 standard phenol mixture eluted in a mobile phase of 1% (v/v) methanol in CO2. The top chromatogram was obtained by averaging currents over a potential range from -0.9 to -1.2 V. The bottom chromatogram was obtained by averaging currents over a potential range from 1.8 to 2 V. The same elution conditions of temperature and pressure as in Figure 1 were used. The last peak to elute, 4-nitrophenol, is labeled. Table 1. Quantitative Aspects of Phenol Detection

compound phenol 2-chloro 2,4-dichloro 4-chloro-3-methyl 2,4-dimethyl 2-methyl-4,6-dinitro 2,4-dinitroc 2,4,6-trichloro 2-nitrophenol pentachloro

peak ht sensitivitya (nA/ng)

est detectn limit (ng)

EC

FID

0.0541 0.0239 0.0229 0.0204 0.0196 0.0109

0.72 1.6 1.7 1.9 2.0 3.3

0.9989 0.9979 0.9963 0.9972 0.9933 0.9968

0.9997

0.0050 0.0049 0.0005

7.8 7.3 80

r2 b

0.9993 0.9936

0.9998 0.9990 0.9993 0.9983 0.9992 0.9993

a Peak height sensitivities were taken from the slope of the best-fit regression line of the calibration data. b Regression coefficients were calculated when at least four data points could be plotted from the calibration data. c 2,4-Dinitrophenol peak heights were difficult to measure at low levels for unknown reasons, regardless of the detector. As indicated by Figure 1, the electrochemical response to 2,4dinitrophenol at high concentrations is similar to that of 2-methyl4,6-dinitrophenol.

Quantitative Aspects of Voltammetric Detection of Phenols in Unmodified CO2. Table 1 provides a quantitative comparison of the performance of the electrochemical detector to that of the downstream FID for 10 of the 11 phenols in the mixture (4-nitrophenol was omitted from this study because of its excessive retention time). Calibration curves were generated by injecting a series of diluted phenol mixtures containing 1-500 ng of each of the phenols. Best fits to the calibration data for each phenol were obtained by linear regression.

The first column in Table 1 gives the regression slopes of calibration curves for the individual phenols. It is interesting to take note of some trends in the variation of sensitivity toward the oxidizable and reducible phenols as a function of their substituents. The sensitivity toward chloro-substituted phenols, for example, decreases as the number of chlorine substituents increases. The sensitivity for pentachlorophenol, the only fullysubstituted phenol in the mixture, is 10% of that toward 2,4,6trichlorophenol. Conversely, the sensitivity for the reducible nitrophenols increases upon the addition of a second nitro substituent. The sensitivity to 2-methyl-4,6-dinitrophenol is approximately twice as large as the sensitivity to 2-nitrophenol. The changes in electrochemical sensitivity that result from changes in the substituents on phenol have at least two possible origins. First, structural changes affect the permeability (i.e., the partitioning ratio of analytes between CO2 and the PEO-Li film and the mass transport rate of analytes in the film) of the individual phenols. Second, structural changes affect the redox half-wave potentials. The addition of a nitro substituent to the para position of phenol, for example, makes the compound impossible to oxidize in the potential range used in this work. The oxidation potential is shifted by the electron-withdrawing nature of nitro substituents. On the other hand, the nitro substituent shifts the reduction potential sufficiently that the compound can be detected by reduction within the potential range used here. These trends in the electrochemical response with changing substitution patterns are interesting because they demonstrate that electrochemical behavior of the phenols under SFC conditions is quite similar to that observed in more conventional electrochemical solvents. The second column in Table 1 reports the detection limit for each phenol. These detection limits were obtained by extrapolating the best-fit calibration line to a current equal to three times the standard deviation of the baseline noise. In all cases, the reported detection limit is at least 75% of the smallest quantity actually detected in these experiments. The detection limits are in the low-nanogram range and are substantially lower than those obtained with the FID. Figure 4 shows chromatograms obtained simultaneously with the electrochemical detector (A) and the FID (B) during the elution of a mixture containing 1-5 ng each of the phenols. The electrochemical peaks for six of the seven oxidizable phenols are well above detection limits, while no peak is observed for the difficult to oxidize pentachlorophenol. On the other hand, at the FID only the peak for 4-chloro-3-methylphenol, which elutes at 14.5 min, is above detection limits. The signalto-noise ratios calculated from the data for 4-chloro-3-methylphenol in Figure 4 are 30 and 5 for the electrochemical detector and the FID, respectively. The noise was determined by calculating the standard deviation of the baseline signals during the 1 min just before the peak. Figure 4 clearly demonstrates that the electrochemical detector is capable of providing lower detection limits than the FID for these oxidizable phenols. This is significant because there exists a need to detect phenols at low levels.19-22 Table 1 also reports the correlation coefficients of the calibration lines obtained with both the electrochemical detector and the FID. All correlation coefficients are 0.993 or better, with the electrochemical detector exhibiting slightly more scatter. In a separate experiment the reproducibility of the electrochemical response to phenols was determined by six replicate injections of a mixture containing 9.9 ng of 2,4-dimethylphenol, 9.85 ng of phenol, and 48.9 ng of 4-chloro-3-methylphenol. The relative

Figure 4. Chromatograms of a 500×-diluted EPA 604 standard phenol mixture obtained with the electrochemical detector (A) and the FID (B). Amounts of the individual phenols injected range from 1 to 5 ng.

standard deviations (RSDs) for the peak heights of each compound were 6.95, 13.5, and 2.20%, respectively. The slightly poorer correlation coefficients and RSDs obtained with the electrochemical detector are the result of a slow decrease in electrode sensitivity over time (see discussion of stability, below). Despite the known tendency of phenols to passivate electrode surfaces, the electrochemical detector response is linear. Long-Term Stability of the Electrochemical Detector. The response of chromatographic detectors used for routine analysis must be stable for long periods of time and must be easy to maintain. To evaluate the capacity of the electrochemical detector for long-term use, a single electrode was used over a 3-day period without being removed from the detector compartment. During each day, the detector was used heavily in order to obtain calibration data for the phenols. At the end of each day, the detector was stored at 60 °C but the CO2 pressure was not maintained at supercritical conditions overnight because of the limited volume of the syringe pump. Figure 5 shows the calibration slopes and linearities of six oxidizable phenols obtained over the course of this experiment. Figure 5A shows that from day one to day three the slopes of the calibration curves decrease by 36-53%. Figure 5B, however, shows that there is no obvious trend in the correlation coefficients obtained from day to day. This reflects the observation that the change in electrode performance was most significantly affected by overnight storage. This suggests that paying closer attention to the overnight storage conditions may extend the stability of the detector response, though this has not yet been done. Figure 5, therefore, shows that the electrochemical detector can be used for phenol analysis for several days without refurbishing. Nevertheless, Figure 5 also shows that the detector will likely require refurbishing during regular use. The detector was designed with this in mind: removal and replacement of the electrode unit of the detector is a 30 s operation. The electrode unit is mounted in a conventional ferrulled fitting and is replaced simply by unscrewing one unit and screwing in the replacement. Equilibration of the replacement detector is typically very quick Analytical Chemistry, Vol. 68, No. 18, September 15, 1996

3125

Figure 6. Segments of chromatograms obtained with the electrochemical detector (A and B) and an FID (C). (A) was obtained by averaging currents over a potential range from 1.8 to 2 V. (B) was obtained by averaging currents over a potential range from -0.9 to -1.2 V. The numbered peaks correspond to the following phenols: (1) 2-methyl-4,6-dinitrophenol, (2) 2,4-dinitrophenol, (3) 2,4-dichlorophenol, and (4) 2,4,6-trichlorophenol. Figure 5. Bar graphs of the calibration slopes (A) and regression coefficients (B) obtained for six oxidizable phenols. The data for these graphs were collected over three consecutive days with a single electrode. Numbered phenols correspond to the following compounds: (1) 2-chlorophenol, (2) 2,4-dichlorophenol, (3) 2,4,6-trichlorophenol, (4) 2,4-dimethylphenol, (5) phenol, and (6) 4-chloro-3methylphenol.

and occurs in under 15 min. This is less than the time required to refill the syringe pump and reequilibrate the column. So, if the detector removal and replacement is made to coincide with refilling the pump, then refurbishing the detector produces no additional down time. The durability of detector response combined with the easy detector maintainence suggests that electrochemical detection in SFC is a practical and useful technique. Selectivity of Voltammetric Detection of Phenols. One of the advantages associated with voltammetric detection is the selectivity of the applied potential axis. Figure 6 shows three chromatograms obtained from the same injection of a mixture of phenols using different data workup parameters. Traces A and B were obtained with the electrochemical detector by measuring currents from a potential range of 1.8-2.0 V on the positive-going sweep (A) and a potential range of -0.9 to -1.2 V on the negativegoing sweep (B). Figure 6C was obtained with the FID. It is evident from the FID response in Figure 6C that four peaks are present, but peaks 2 and 3 are unresolved. The electrochemical detector is able to resolve these overlapping peaks because peak 2 (2,4-dinitrophenol) is reducible but not oxidizable and peak 3 (2,4-dichlorophenol) is oxidizable but not reducible in the chosen potential windows. In this example, the adjustment of potential between (A) and (B) accomplishes two things: (1) amplification of the electrochemical response for 2-methyl-4,6-dinitrophenol (peak 1) and 2,4-dinitrophenol and (2) resolution of the 2,4dinitrophenol and 2,4-dichlorophenol peaks. Figure 7 demonstrates another voltammetric strategy for increasing detector selectivity. The chromatograms in (A) and (B) were obtained by measuring currents from potential ranges 3126 Analytical Chemistry, Vol. 68, No. 18, September 15, 1996

Figure 7. Segments of chromatograms obtained with the electrochemical detector. (A) was obtained by averaging currents over a potential range from 1.95 to 2 V. (B) was obtained by averaging currents over a potential range from 1.58 to 1.8 V. (C) was obtained by subtracting (B) from (A).

of 1.95-2 V (A) and 1.58-1.8 V (B) on the positive-going sweeps. The chromatogram in (C) was obtained by subtracting the current values in (B) from the current values in (A). The pressure program used in this experiment did not produce complete resolution of all of the phenols. The arrows in Figure 7A point to background-subtracted voltammograms that were collected at the top of the two unresolved peaks. The voltammograms differ sufficiently that the second peak (2,4-dimethylphenol) can be

Figure 8. Successive chromatograms of a 10×-diluted EPA 604 standard phenol mixture obtained with the electrochemical detector operated amperometrically. A constant applied potential of 2.0 V was used. The numbered peaks correspond to the same phenols as in Figure 1. (See Figure 1 legend.)

completely resolved from the first (2,4,6-trichlorophenol) by selecting a lower range of oxidation potentials (Figure 7B). Subtraction of the chromatogram in (B) from the chromatogram in (A) yields a completely resolved 2,4,6-trichlorophenol peak in (C). In a way not possible with an FID, complete resolution of two overlapping peaks has been demonstrated with a single voltammetric detector. In Figure 6, the potentials used to resolve overlapping peaks differed by nearly 3 V. In Figure 7, however, resolution of overlapping peaks was accomplished for potentials that differ by only 0.2 V. Amperometric Detection of Phenols in Unmodified CO2. So far we have discussed only voltammetric detection in CO2. The question arises as to whether or not constant-potential amperometric detection, which is technically simpler than voltammetric detection, offers any value to SFC. Figure 8 shows segments of two chromatograms obtained in a pilot study of amperometric detection for successive replicate injections of 11 priority phenols in unmodified CO2. A constant potential of +2 V vs Pt QRE was maintained during each injection. The numbers on each peak are

consistent with the numbering scheme in Figure 1. The chromatograms in Figure 8 show that amperometric detection is unfit for use in CO2. Midway through the chromatogram in Figure 8A, which was the second chromatogram to be obtained with this detector, a sudden drop in the baseline current occurred. Subsequent peaks in the remainder of the chromatogram and in the next chromatogram, shown in Figure 8B, have greatly decreased sensitivities. Thus, midway through the second chromatogram generated by amperometric detection, the electrode failed catastrophically. Amperometric detection is not fit for SFC. Compared to amperometric detection, therefore, voltammetric detection is far superior. There are at least two factors that contribute to the superior performance of voltammetric detection. First, sweeping the potential shortens the length of time spent oxidizing or reducing eluents, relative to constant potential amperometry. At a sweep rate of 25 V/s and an interval between scans of 2 s, the working electrode is poised at phenol oxidation potentials (1.8-2 V) only 0.8% of the time. As a result, smaller amounts of electrogenerated products are formed. If the oxidation products passivate electrodes, which is known to occur in the case of phenols, then the extent of passivation should be less severe when voltammetry is used. Second, holding the potential constant for long periods of time places a strain on the so-called ion budget of the self-contained electrochemical cell. Sweeping the potential, which causes currents to flow in both directions through the PEO-Li film, diminishes that strain. Conclusion. The results of this work demonstrate that voltammetric detection for SFC is applicable to a broad range of compounds, many of which must be routinely monitored in the environment. The compactness and instrumental simplicity of the voltammetric detector is a distinct advantage that makes it suitable for use in on-site, mobile laboratories. Furthermore, the selectivity of voltammetric detection provides added resolution of overlapping peaks and chemical information regarding the nature of the eluting compounds. ACKNOWLEDGMENT This work was supported by the Central Research and Development Fund of the University of Pittsburgh.

Received for review December 7, 1995. Accepted June 17, 1996.X AC951188L X

Abstract published in Advance ACS Abstracts, August 1, 1996.

Analytical Chemistry, Vol. 68, No. 18, September 15, 1996

3127