ANALYST, JANUARY 1993, VOL. 118 17 Supercritical Fluid Extraction for the Analysis of Liquid Poly(alky1ene glycol) Lubricants and Sorbitan Ester Formulations Terence P. Hunt and Chris J. Dowle* ICI Wilton Research Centre, P.O. Box 90, Middlesbrough, Cleveland, UK TS6 8JE Gillian Greenway Department of Chemistry, University of Hull, Hull, UK HU6 7RX Additives were selectively extracted from poly(al kylene glycol) (PAG) lubricant formulations using supercritical fluid extraction. The PAG matrix was adsorbed onto silica for the extractions and the extracted analytes were passed through a silica column positioned in-line in the supercritical fluid stream. The selectivity obtained by this method was compared with that obtained by the direct extraction of adsorbed and unadsorbed PAG and by the extraction of unadsorbed PAG through the in-line column.The final procedure was found t o be successful in separating additives from all but the lowest molecular mass PAG oligomers. Further use of this procedure was demonstrated by a stepwise extraction of a sorbitan ester formulation. This gave a chemical-class fractionation of the product and so provided a sample preparation technique for further spectroscopic analysis. Keywords: Supercritical fluid extraction; liquid poly(alky1ene glycol) and sorbitan ester samples; adsorption on silica; in-line silica packed column Supercritical fluid extraction (SFE) has been used extensively in the analysis of solid polymers. 1,2 Supercritical fluid extrac- tion of liquid samples is undertaken less widely because dissolution or entrainment of the matrix can occur.Direct extraction of aqueous samples has resulted in incomplete separations3.4 and although a more promising extractor, incorporating a dip tube, has been used by Ong et aZ.5 to extract cholesterol from egg yolk, this system was not used for more volatile samples. An alternative approach , analogous to thermal desorption in gas chromatography, involves immob- ilizing a sample on an adsorbent bed which is then extracted by a supercritical fluid, Polycyclic aromatic hydrocarbons, chlori- nated hydrocarbons, chlorinated aromatics and organophosp- hate pesticides have been extracted from Tenax resin.6.7 Moreover, Wright et a1.8 using several different adsorbents and extraction fluids showed that the selective recovery between individual solutes can be varied by altering the combination of the adsorbent and the extraction conditions.In this way Alexandrou et al.9 successfully separated poly- chlorinated dibenzo-p-dioxins (PCDDs) from polychlorinated biphenyls and chlorinated benzenes in standard and fly ash extract solutions. The bulk of the solutes were extracted from Florisil with C 0 2 at 20.68 MPa after which the remaining PCDDs were extracted with N20 at 41.37 MPa. Selectivity in the extraction process can also be achieved by passing the effluent from the extraction cell through an adsorbent column. This procedure was used to remove contaminants from seed oil extract by King et a1.10 using an activated carbon column and by Shishikura et aZ.11 to remove cholesterol from butter oil extract using a silica column.Ramsey et aZ.12 cleaned up freeze-dried pig's kidney extracts for on-line drug analysis by mass spectrometry (MS). The polar drugs were adsorbed onto an amino column while the bulk of the non-polar endogenous components were extracted rapidly to waste. The addition of a polar modifier to the supercritical fluid then enabled the drugs to be desorbed to the MS system. Also, Yamauchi et al.13 extracted naphthalene selectively with supercritical C02 at 10 MPa from a mixture of naphthalene and anthracene by passing the extracted solutes through a silica column. Naphthalene eluted through the column but anthracene was adsorbed. The anthracene was then desorbed by increasing the pressure to 25 MPa . * To whom correspondence should be addressed. Poly(alky1ene glycols) (PAGs) are low molecular mass liquid polymers which consist of chains of ethylene oxide (EO) and propylene oxide (PO) units bonded to an alkoxy end-group.The analysis of PAG additives (antioxidants, biocides and anti-corrosion , anti-wear and anti-foaming agents) is hindered by the presence of the PAG matrix and so a method for separating additives from PAG is required. However, both PAG and additives tend to be soluble in supercritical C02. Therefore, the selective extraction of additives from PAG must be achieved by immobilizing PAG on an adsorbent in either of the ways described above. Silica is a powerful adsorbent, acting through hydrogen-bonding with surface silanol groups. It has also been found that for solutes containing nitrogen and oxygen, this mechanism is sup- plemented by coordinate bonding with electron acceptor centres on the surface.*4?15 It is a matrix which should, therefore, be suitable for adsorbing EO-PO polymers.The strong interaction of silica with EO-PO polymers is demon- strated by its use as the stationary phase for the normal phase high-performance liquid chromatography (HPLC) separation of oligomers of ethoxylated acids and alcohols. Furthermore, many workers have chosen to derivatize ethoxylates in order to reduce the strength of the interaction which would otherwise be too strong for the elution of higher oligo- mers.16.17 Similar reasoning suggests that silica could also be used as an adsorbent in the selective extraction of formulations of sorbitan ester (a sugar formed by reaction of sorbitol and lauric acid).Sorbitan mono-ester contains three hydroxyl groups and one cyclic oxygen atom in addition to the ester carbonyl group. These oxygen atoms, and particularly the carbonyl , should exhibit strong hydrogen-bonding interac- tions with silica. This seems to be confirmed by reported normal-phase HPLC and thin-layer chromatography (TLC) separations of other sugars on silica stationary phases which require very polar eluents.18.19 In this paper silica has been used both to adsorb PAG and sorbitan ester samples prior to extraction and to fractionate the extracted solutes in the extraction effluent downstream from the extraction cell. The method has been scaled up to enable 1 g samples to be analysed.This provided fractions of extracted analytes in sufficient amounts to be characterized by nuclear magnetic resonance (NMR) spectroscopy without the interferences from solvents or their impurities frequently found with traditional sample preparation schemes.ANALYST. JANUARY 1993. VOL. 118 Column Fig. 1 back-prcssure regulator Schematic diagram o f thc SFE system. PU = Pump; PH = prc-heating column; SV = switching valve; PM = pressure mcter and BR = Experimental Reagents Extractions were performed on PAG lubricants (Union Carbide, ICI and Exxon) and on sorbitan ester (ICI). Carbon dioxide (SFC grade, Air Products) was used for SFE and capillary supercritical tluid chromatography (CSFC). Samples were adsorbed on 5 pm particle size silica Spherisorb S5W and alumina Spherisorb A5Y (Hichrom).Dichloromethane of HPLC grade was used to slurry samples with the oxides, HPLC grade methanol was used as a polar modifier to the C02 and stabilized tetrahydrofuran (THF) was used as the eluent in the gel-permeation chromatography (GPC j analysis of extracts. All solvents were supplied by either BDH or Fisons Laboratory Supplies. Instrumentation Extractions were carried out using the Jasco (SFC-SFE) system shown in Fig. 1 with a 300 x 7.5 mm column case (Polymer Laboratories) used for the extraction cell. This was positioned in the loop enclosed by the Rheodyne 7000 switching valve SV1 and was heated in the 860-CO oven. A 4.6 mm i.d. adsorption column (Spherisorb S5W or Nucleosil silica, all supplied by Hichrom) was positioned in the loop enclosed by the switching valve SV2, downstream from the extraction cell and outside the oven.Adsorption columns were 250 nim long unless otherwise stated. The switching valves SV1 and SV2 allowed the CO? flow to be passed through either the extraction cell and column or by-passed to the 880-81 back-pressure regulator BR2 via the 875-UV (ultraviolet) detector. A third switching valve SV3 was always by-passed during extractions and the Rheodyne 7037 pres- sure-relief valve BRl was adjusted to give negligible back- pressure and the extraction back-pressure was maintained using a Jasco back-pressure regulator BR2. Samples were analysed by CSFC and GPC. The CSFC system consisted of a Carlo Erba SFC 3000 instrument with a 10 m x 50 pm i.d. biphenyl polysiloxane column (Dionex, UK).The GPC system consisted of a Knauer pump and refractive index (RI) detector, a Pye Unicam temperature controller and a Talbot ASI-3 aut?sampler with a 300 X 7.5 mm 104 8, + 300 x 7.5 mm 500 A column series (Polymer Laboratories). Table 1 Extraction conditions Extraction Back- In-line scheme Methanol pressure Flow ratc adsorption number -COz (%) /MPa Time/min /ml min-1 column 1 0 20 30 1 .o No 0 30 30 1 .0 2 0 30 1380-900 1.0 Y cs 3 0 30 240 4.0 Yes 4 0 30 60 1 .o 5 0 20 I000 1 .0 Yes 6 0 10 60 4.0 Y cs 0 20 1000 I .0 1 0 30 60 4.0 40 30 60 4.0 0 7 90 4.0 0 10 280 4.0 0 15 70 4.0 0 20 70 4.0 0 30 990 1 .0 10 30 70 4.0 40 30 150 4.0 7 Y es Procedure All extractions were carried out at 45 "C with the adsorption column (when used) at ambient temperature.The UV absorbance of extracted solutes was monitored with the detector at 220 nm and extracts were collected in glass collection vessels positioned beneath BR2. The extraction conditions used are listed in Table 1 and after each extraction the SFE system was cleaned by pumping through with 40% MeOH-C02 at 30 MPa. Adsorbed samples were prepared by slurrying a 1 : 5 sample : oxide mixture in dichloromethane and evaporating to dryness. Initially the extraction procedure was optimized for PAG A to determine how sample adsorption onto silica and extraction through an in-line silica adsorption column affects extraction selectivity. To achieve this the extraction selectivity was compared for extractions on free (unadsorbed) PAG A and a PAG A-silica mixture without an in-line adsorption column and for PAG A extractions through in-line adsorption columns packed with silicas of various pore sizes.ExtractionsANALYST, JANUARY 1993, VOL. 118 19 with an in-line adsorption column were then compared for free PAG A, PAG A-silica and PAC A-alumina to investigate how the combination of sample adsorption and an in-line Extractions on Unadsorbed Samples With In-line Adsorption Column column affects the extraction selectivity. Thus free PAC A and a PAGA-silica mixture were each directly extracted with SV2 closed at 20 and 30 MPa using extraction scheme 1 (see Table 1). Free PAG A was also extracted overnight at 30 MPa and 1 ml min-1 with SV2 open so that the CO? flow and extracted solutes were passed through a Spherisorb column (extraction scheme 2).This was repeated for :00 mm Nucleosil columns with pore sizes of 50, 120 and 300 A. PAC A was also extracted through a 50 mm Spherisorb column at 4 ml min-1 (extraction scheme 3). PAG A and mixtures of 0.1 : 0.5 g PAG A-silica and PAG A-alumina were then extracted at 30 MPa for 60 min (extraction scheme 4) through a Spherisorb column and these were compared with an identical PAC A-silica extraction in which SV2 was closed to by-pass the column. The final, optimized method was then used to extract various samples. Hence 1 : 5 g PAG A-silica mixtures were extracted overnight through a Spherisorb column (extraction scheme 2). Further extractions through a Spherisorb column were carried out on 1 : 5 g mixtures of PAGs B and C and sorbitan ester with silica. PAG B was extracted at 20 MPa (extraction scheme 5 ) ; PAC C and sorbitan ester were extracted stepwise from 10 and 7 MPa (CO,) to 30 MPa (40% methanol-C02) (extraction schemes 6 and 7 respectively).PAGs A and B and their extracts were analysed by GPC with a stabilized THF eluent at 1 ml min-1 and ambient temperature. PAG extracts shown by GPC not to contain high molecular mass material were also analysed by CSFC at 100 "C using a C 0 2 pressure gradient of 6 or 1W.5 MPa at 1 MPa min-1. PAG C and sorbitan ester and their extracts were analysed by CSFC: with a temperature gradient o f 100-190 "C at 3 "C min-1 and an asymptotic density gradient of 0.16-0.5 over 30 min (PAG C); at 100°C with an asymptotic C 0 2 density gradient of 0.15-0.55 g ml-1 over 40 min (sorbitan ester).Results and Discussion Direct Extractions (Without In-line Adsorption Column) PAG A and the 30 MPa SFE direct extracts of PAG A and PAG A-silica were analysed by GPC to give a measure of the concentrations of additive. The concentration of additive is proportional t o its GPC peak area expressed as a percentage of the total peak area of the sample (referred to as relative peak area in this paper). These results are presented in Table 2. Although PAG is present in both extracts, comparison of the additive relative peak areas listed in Table 2 shows that the additive concentration is greater in the PAG A-silica extract. This shows that by interacting PAG with silica, the selective extraction of more weakly adsorbed additives is enhanced, although insufficiently to provide a complete separation of the additives from PAG.Table 2 Rclative GPC peak area of additive (total peak area set to 100) for PAG A and PAG A extracts Relative peak area of Sample additive Unextracted PAG A- 7 6 CO? PAG A-silica direct extraction 44 PAG A through-column extraction 29 PAG A-silica through-column 100 30 MPa 1 PAG A dircct extraction extraction Rather than slurrying samples with silica to give adsorbed mixtures prior to extraction it would be simpler to extract the free sample and pass the extracted solutes through a silica column. For PAG A this appears to be successful for relatively short extraction times. After 1-2 h the extract collected from a 1 ml min-' 30 MPa C 0 2 extraction through a silica adsorption column, is a creamy white solid. However, when the extrac- tion is continued overnight for 315 h, the extract becomes an oil resembling the original sample.Moreover, even after 1000 min, the UV absorbance of the extraction effluent from the adsorption column shows no sign of decaying to the pre- extraction value (set to zero). This shows that the additives are not completely extracted and suggests that a final extract would be even further contaminated with PAG. In order to run such an extraction to completion, PAG A was extracted at 4.0 ml min-1 with a shorter SO mm in-line adsorption column. The extract was analysed by GPC for the additive relative peak area which is listed in Table 2. This shows that greater PAG contamination results from this method than for the direct extraction of PAG A-silica. However, enhanced selectivity over the direct extraction of unadsorbed PAG A is achieved.The extraction selectivity can be affected by changing the properties of the adsorption column silica. The UV absorb- ance curves for CO, extractionsoof PAG A through silicas with pore sizes between 50 and 300 A are shown in Fig. 2. It can be seen from this that the extraction rate increases with pore size. This is consistent with a smaller surface area available to adsorb the extracted sample resulting from a larger pore size. As extracts become more contaminated with PAG the further the extraction progresses it might be expected that the lowest concentration of PAG would be found in the 50 A extract. However, the results ot GPC analysis, given in Table 3, show that it is in fact the 120 A extract that is the least contaminated.A possible explanation is that a larger pore size allows the large PAC molecules to enter the silica pores more easily. This should enhance PAG-silica adsorption and improve the extraction selectivity. Owing to these two conflicting tenden- cies an optimumo extraction selectivity is obtained at the intermediate 120 A pore size. Extractions on Adsorbed Samples With In-line Adsorption Column The UV absorbance curves for various 60 min extractions of 0.1 g samples of PAG A with 30 MPa C 0 2 (extraction scheme 4) are shown in Fig. 3. In each instance the sample size and extraction conditions are identical. They show the direct 0 200 400 600 800 Time/min Fig. 2 UV absorbance of 30 MPa C02 extractions of PAG A through 100 mm silica columns with different pore sizes: A, 50; B, 120; and C , 300 A.(Extraction scheme 2)20 ANALYST, JANUARY 1993, VOL. 118 Table 3 Relative GPC peak area of additive for 30 MPa C02 extracts from extractions of unadsorbed PAG A through silica columns of different pore size Pore sizciA Relative peak area of additive 50 120 300 36 56 30 z 600 . 4- x cn .- 2 400 - 0 10 20 30 40 50 Time/min Fig. 3 UV absorbance of 30 MPa COz extractions of: A, 1 : 5 PAG A-silica; B, PAG A through a 250 mm silica column: C, 1 : 5 PAG A-alumina through a 250 mm silica column; and D, 1 : 5 PAG A-silica through a 250 mm silica column. (Extraction scheme 4) extraction of PAG A-silica [extraction (a)] and the through- column extractions of unadsorbed PAG A, PAG A-alumina and PAG A-silica [extractions ( h ) , ( c ) and (d), respectively].Comparison of extraction curves in Fig. 3 shows a marked improvement in the definition of individual peaks for extrac- tion (d) where adsorption of the sample prior to extraction and passing the extracted solutes through an adsorption column are combined. The difference between extractions (b) and (d) cannot be explained by the extra mass of silica (0.5 g) mixed with the sample prior to extraction as this is insignificant compared with the mass of silica contained in the adsorption column. This suggests that adsorption onto the dry silica used in premixing the sample is much stronger than adsorption onto the silica contained in the column in the supercritical fluid stream.It is known from studies in normal-phase HPLC that even non-polar eluents such as heptane are weakly adsorbed on the surface of a silica column to form a monolayer coating of solvent molecules.20 A similar monolayer formed by supercritical C 0 2 on a silica column would force extracted solute molecules to compete with the C02 molecules in this monolayer in order to be adsorbed. This would account for the weaker adsorption. The difference in resolution between the silica and alumina adsorbed samples shows the importance of choosing the right adsorbent in optimizing the extraction selectivity . When a scaled up 1 : 5 g PAC A-silica through-column C 0 2 extraction is run overnight, the UV absorbance curve of the extracted additive shows that the extraction is now completed rapidly after about 600 min whereas the equivalent extraction of unadsorbed silica under identical conditions is still far from complete after 1000 min.Analysis of the PAG A-silica extract by GPC and CSFC shows that it is free of PAG contamination (see Table 2). Therefore, in this extraction additives are selectively extracted from PAG. Similar results are obtained for the 20 MPa through-column extraction of PAG B. The UV absorbance curve, shown in Fig. 4, shows that the PAG B additives are extracted from the silica column in two clearly resolved fractions which were collected separately. The second peak was confirmed as 1000 800 > @ 600 2- cn 4- .- 400 - 200 0 100 200 300 400 500 600 Time/min Fig. 4 UV absorbance of a 20 MPa C02 extraction of 1 : 5 PAG B-silica through a 250 mm silica column (extraction scheme 5 ) Table 4 Molecular mass values for PAG A and PAG B Mass-average Number-average Molecular mass molecular mass molecular mass at peak maximum PAG A 2770 2610 2750 PAG B 3370 2290 3190 z 18.0 -.14.0 12.0 0 lrganox 1010 THF stabilizer 6.0 12.0 18.0 24.0 30.0 36.0 42.0 Tim e/m i n Fig. 5 CSFC trace of 1 : 5 PAG B-silica 20 MPa C02 extract passed through a 250 mm silica column (extraction scheme 5) Irganox 1010 antioxidant by CSFC and NMR. The GPC analysis shows that this Irganox 1010 fraction of the 20 MPa extract is separated from the bulk of the PAG. This result should be expected as GPC analysis gives similar molecular mass values for both PAGs (see Table 4). However, PAG B contains lower molecular mass oligomers not present in PAG A and the CSFC trace of the 20 MPa extract shows that some of these shorter-chain oligomers have been extracted together with Irganox 1010 (see Fig.5 ) . Clearly the higher the molecular mass of a PAG chain, then the more EO-PO groups in the chain are available to be adsorbed onto the silica and consequently the over-all adsorption of a PAG chain is stronger. Fig. 5 shows that a threshold number of groups is needed for a PAG chain to be sufficiently adsorbed for a given set of extraction conditions, with PAG chains below this threshold being extracted. This effect is better illustrated for the stepwise, through- column extraction of PAG C-silica. PAG C has a nominal molecular mass of only 750 and CSFC analysis of its extracts, shown in Fig.6, shows that the first 9-10 oligomers can be extracted by 10 MPa C02. Increasing the extraction pressure to 20 MPa boosts the C 0 2 solvent strength and so allows the further desorption of oligomers 10-15 whilst the remaining longer-chain oligomers are desorbed when the extraction fluid polarity is increased by the addition of methanol. Clearly PAGANALYST, JANUARY 1993, VOL. 118 21 adsorption must be strengthened in order to separate additives from shorter-chain oligomers. Further studies are planned to investigate the possibility of achieving this either by using a different silica or even by changing to a polar, bonded phase such as a diol or amine. Extractions on Sorbitan Ester Having developed the above method for the separation of additives from PAG formulations, the possibility of using it in the analysis of different samples was then considered.Sorbitan ester is a complex mixture of sugars formed by the reaction of lauric acid and sorbitol. Both reactants may contain impurities and the final product contains mono-, di- 60 50 > E 40 . c > KI m .- E: 30 - 20 10 0 6.0 12.0 18.0 24.0 30.0 36.0 42.0 Time/min Fig. 6 CSFC traces of 1 : 5 PAG C-silica extracts assed through a 250 mm silica column. A, C02 (10 MPa); B, COz 6 0 MPa); and C, 10% mcthanol-C02 (30 MPa). (Extraction scheme 6 ) 24.0 26'o 1 I =. 22.0 E 2 20.0 5 18.0 5 16.0 . .- fn CI 14.0 12.0 19.0 18.0 17.0 16.0 15.0 14.0 13.0 12.0 and tricyclic esters and linear esters. Its CSFC trace is shown in Fig. 7(a). Hence for this sample, the extraction of parts of the sample matrix (as observed in the extraction of PAG C, where it is unwelcome as the aim is to isolate certain trace additives) can be used here to investigate the composition of the matrix itself by providing a potential means of separating it into constituent fractions that are easier to analyse. The stepwise, through-column extraction of this sample (adsorbed onto silica) gives a series of separately collected fractions.The CSFC traces of the main fractions are shown in Fig. 7(b)-V). Comparison with that of the original sample in Fig. 7 ( a ) shows that their compositions are much simpler, which permits less complex spectroscopic analysis to be d'eveloped for different fractions. Initial analysis by 13C NMR suggests that the 7 and 10 MPa COz extracts contain cyclic mono-esters; the 1.5 and 20 MPa C 0 2 extracts contain cyclic di- and tri-esters with some fatty acid (more pronounced in the 20 MPa extract); and the C02 and 10% methanol-C02 extracts contain linear esters.The spectra of the extracted esters were predicted with the assistance of our corporate 13C NMR computerized chemical-shift calculation facility and spectral database. The predictions were confirmed by the subsequent synthesis of the proposed compounds from isomerically pure starting materials. These results confirm that adsorption is enhanced by the addition of further ester carbonyl groups. Also, the greater adsorption of linear compared with cyclic esters may be because they are less sterically hindered in their ability to attach themselves to silica through their oxygen atoms.The 40% methanol-C02 fraction has not been identified. This characterization would be extremely difficult for the original sample. Furthermore, this method creates a separation on the basis of a specific property of the matrix components (their adsorption strength) and should be seen as complementary to 27.0 I 24.0 1 I 15.0 12.0 9.0 11.4 11.2 11.0 10.8 10.6 10.4 10.2 10.0 9.8 14.5 14.0 L I (f' I 12.0 18.0 24.0 30.0 36.0 42.0 48.0 54.0 Time/min 13.5 13.0 12.5 12.0 12.0 18.0 24.0 30.0 36.0 42.0 48.0 54.0 Fig. 7 ( c ) 10 MPa C02 extract; (d) 20 MPa C02 extract; (e) 30 MPa C02 extract; and (f) 30 MPa 40% mcthanol-CO2 extract CSFC traces of sorbitan cstcr and 1 : 5 ester-silica extracts (extraction scheme 7).( a ) Original sample; ( b ) 7 MPa C 0 2 extract;22 ANALYST, JANUARY 1993, VOL. 118 other fractionation techniques, such GPC, which utilize different properties as semi-preparative Conclusions Additives in PAG lubricants cannot be isolated by SFE carried out on free samples because of the relatively high solubilities of both PAG lubricants and their additives in supercritical CO2. The separation is improved by extracting the lubricant sample as a PAG-silica mixture and by passing the effluent from the extraction cell through a silica column. In both instances the sample is adsorbed onto the surface of the silica where additives tend to be more easily desorbed than PAG. This is confirmed by the GPC analysis of the C02 extracts, which shows that the relative areas of additive peaks are greatly increased compared with that for the unextracted sample.When these two procedures are combined so that a PAG-silica mixture is extracted with the extracted solutes passing through a silica column, it is possible to extract less polar additives from PAG exhaustively with only very short PAG chains extracted together with the additives. Further- more, individual additives are desorbed separately in discrete bands which can be collected separately for further analysis. The further application of this method to sorbitan ester resulted in its fractionation into four distinct chemical classes: cyclic monoesters, cyclic di- and tri-esters, linear esters and a highly polar but unidentified fraction. This particular example showed the value of SFE as a sample preparation technique for the spectroscopic characterization of complex mixtures.Thanks to A. Bunn and H. Yates of ICI Wilton Research Centre for NMR analyses. References 1 Cotton, N . J . , Bartle, K. D., Clifford, A . A., Ashraf, S . , Moulder, R., and Dowle, C. J . , J. High Resolut. Chromatogr., 1991, 14, 165. 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Hunt, T. P., Dowle, C. J., and Greenway, G., Analyst, 1991, 116, 1299. Roop, R. K., Hess, R. K., and Akgerman, A.. Supercrit. Fluid Extr. Chromatogr. (ACS Symp. Ser.), 1989. Tiebault, D., Chervet, J. P., Vannoort, R. W., de Jong, G. J., Brinkman, U. A. Th., and Frei, R. W., J . Chromatogr., 1989, 477, 151. Ong, C. P., Ong, H. M., and Li, S. F. Y . , J. Microcol. Sep., 1990, 2, 69. Hawthorne, S. B., and Miller, D. J., J. Chromatogr. Sci., 1986, 24, 258. Raymer, J . H., and Pcllizzari, E. D., Arzal. Chem., 1987, 59, 1043. Wright, R. W., Wright, C. W., Gale, R. W., and Smith, R. D., Anal. Chem., 1987, 59, 38. Alexandrou, N., Lawrence. M. J . , and Pawliszyn, J . , Anal. Chem., 1992, 64, 301. King, J. W., Eissler, R. L., and Friedrich, J . P., Supercrit. Fluid Extr. Chromatogr. (ACS Symp. Ser.), 1988, 63. Shishikura, A., Fujimoto, K., Kaneda, T., Arai, K., and Saito, S . , Agric. Biol. Chem., 1986, 50, 1209. Ramsey, E. D., Perkins, J. R., Games, D . E., and Startin, J . R . , J. Chromatogr., 1989, 464. 353. Yamauchi, Y., Kuwajima, M., and Saito, M., J. Chromatogr., 1990, 515,285. Tret’yakov, N. E . , and Filimonov, V. N., Kinet. Katal., 1972, 13, 815. Filiminov, V. N . , Lopatin, Y. N., and Sukhov, D . A., Kinet. Katal., 1969, 10, 458. Aitzmuller, K., J. Chromatogr. Sci., 1975, 13, 454. McClure, J. D., J. Am. Oil Chem. SOC., 1982, 59, 364. Karamanos, N. K., Tsegenidis, T., and Antonpoulos, C. A., J. Chromatogr., 1987, 405, 221. Koizumi, K., Utamura, T., and Okada, Y., J. Chromatogr., 1985, 321, 145. Brown, P. R., and Hartwick, R. A., High Performance Liquid Chromatography, Wiley-Interscience, New York, 1989. Paper 2103926E Received July 22, 1992 Accepted October 12, 1992