Anal. Chem. 1997, 69, 1612-1619
Equilibrium and Rate Study of Analyte-Matrix Interactions in Supercritical Fluid Extraction Thomas M. Young* and Walter J. Weber, Jr.
Civil and Environmental Engineering, University of Michigan, Ann Arbor, Michigan 48109
Variation in analyte extraction efficiencies from environmental samples using supercritical carbon dioxide (SC CO2) have often been attributed to “matrix effects” including strong matrix-analyte binding. To examine this hypothesis, the equilibrium distribution of phenanthrene between SC CO2 and five well-characterized natural materials was measured as a function of temperature, pressure, and methanol modifier concentration. Dynamic extraction efficiencies and rates with and without methanol modifier were measured from the same materials. Energetic heterogeneity of active sites on dry soils was suggested by nonlinear isotherms displaying significant capacity differences across soil types. Removal efficiencies were highest from low organic matter sorbents, and these materials exhibited the lowest phenanthrene capacities and most linear isotherms. Modifier addition decreased the sorbent’s equilibrium capacity for phenanthrene sorption by as much as an order of magnitude, resulting in linear isotherms and offering strong evidence that methanol acts by competitively displacing analyte from high-energy sites. Temperature increases were more effective than pressure increases in stimulating desorption from soils, consistent with previously observed temperature effects in supercritical fluid extraction. A nonlinear local equilibrium model including no mass transfer limitations was employed to illustrate the impact of variations in isotherm linearity and capacity on the extraction process. Analytical supercritical fluid extraction (SFE) has been applied to an expanding array of analytes and matrices, and standardized approaches have begun to emerge.1-5 Carbon dioxide is the most attractive and widely used supercritical solvent because it is environmentally benign and offers an accessible critical region. Pure supercritical CO2 (SC CO2) is frequently unable to quantitatively remove analytes from real-world environmental samples, even under conditions producing high target compound solubilities, a phenomenon referred to as the matrix effect. Although the performance of all extraction methodologies (e.g., Soxhlet extraction, sonication, accelerated solvent extraction) may be
limited by matrix effects to varying degrees, supercritical fluids offer a unique opportunity to control the solvating power of the extraction fluid by varying temperature, pressure, or cosolvent content to probe matrix-solute interactions. For example, recent studies have shown that matrix effects in SFE can be overcome by judicious selection of cosolvent and/or by elevating extraction temperatures.6-8 However, the exact nature of the matrix effect and its relation to natural sorbent characteristics remain largely unknown. Studies that have examined extraction efficiency differences between sorbents have typically employed synthetic sorbents or natural sorbents for which few details of physical/chemical characteristics are provided.2,9-11 An important result of this previous work is that analytes extracted quantitatively from one natural or synthetic matrix at a particular supercritical fluid condition may not be mobilized appreciably from another under the same conditions. Several hypotheses have been advanced regarding sorbent characteristics that determine extraction success, including clay content, organic matter content, moisture content, and particle size, but few conclusions have been reached. The efficiency enhancement resulting from addition of small amounts of cosolvent to SC CO2 has been explained by increased solubility (analyte-modifier interactions), displacement of analyte from sorption sites on the solid (modifier-matrix interactions), or increased extraction rates (matrix swelling). Cosolvents can greatly enhance organic chemical solubility in SC CO2 by clustering about solubilized molecules, stabilizing them through dispersive or specific chemical interactions.12-14 A portion of the modifier effect undoubtedly results from such solubility enhancements. However, the marginal reduction in modifier effectiveness at higher modifier concentrations and the dependence of modifier efficacy on matrix characteristics suggest the importance of modifier-matrix interactions.7,8 Modifier effects may also arise because the cosolvent swells the matrix and increases extraction rates.15
* Address correspondence to this author. Present address: Department of Civil and Environmental Engineering, University of California, Davis, CA 95616. Fax: 916-752-7872. E-mail:
[email protected]. (1) Hawthorne, S. B. Anal. Chem. 1990, 62, 633-642. (2) Alexandrou, N.; Lawrence, M. J.; Pawliszyn, J. Anal. Chem. 1992, 64, 301311. (3) Lin, Y.; Smart, N. G.; Wai, C. M. Environ. Sci. Technol. 1995, 29, 27062708. (4) Field, J. A.; Miller, D. J.; Field, T. M.; Hawthorne, S. B.; Giger, W. Anal. Chem. 1992, 64, 3161-3167. (5) McNally, M. E. P. J. AOAC Int. 1996, 79, 380-387.
(6) Langenfeld, J. J.; Hawthorne, S. B.; Miller, D. J.; Pawliszyn, J. Anal. Chem. 1993, 65, 338-344. (7) Langenfeld, J. J.; Hawthorne, S. B.; Miller, D. J.; Pawliszyn, J. Anal. Chem. 1994, 66, 909-916. (8) Yang, Y.; Gharaibeh, A.; Hawthorne, S. B.; Miller, D. J. Anal. Chem. 1995, 67, 641-646. (9) Wright, B. W.; Wright, C. W.; Gale, R. W.; Smith, R. D. Anal. Chem. 1987, 59, 38-44. (10) Hawthorne, S. B.; Miller, D. J. J. Chromatogr. Sci. 1986, 24, 258-263. (11) Hawthorne, S. B.; Miller, D. J. Anal. Chem. 1987, 59, 1705-1708. (12) Dobbs, J. M.; Wong, J. M.; Johnston, K. P. J. Chem. Eng. Data 1986, 31, 303-308. (13) Dobbs, J.; Wong, J. M.; Lahiere, R.; Johnston, K. P. Ind. Eng. Chem. Res. 1987, 26, 56-65. (14) Tomasko, D. L.; Knutson, B. L.; Pouillot, F.; Liotta, C. L.; Eckert, C. A. J. Phys. Chem. 1993, 97, 11823-11834.
1612 Analytical Chemistry, Vol. 69, No. 8, April 15, 1997
S0003-2700(96)01014-1 CCC: $14.00
© 1997 American Chemical Society
The effects of elevated SFE temperatures on extraction efficiency have been the subject of recent investigations. Earlier SFE research focused on maximizing CO2 density by keeping temperatures near the critical point (31.1 °C) and raising pressure, based on correlations between density and analyte solubility.11 Although solubility is enhanced by reducing temperature within the “retrograde region”, at typical SFE pressures (e.g., above 200 atm) raising temperature may greatly enhance solubility.16-18 Consistent with these findings, more recent work has shown that large improvements in efficiency are possible by raising extraction temperatures.6,8 In addition to enhancing solubility, high temperatures may assist extraction by overcoming activation energy barriers to desorption from high-energy surface sites.2 Elevating extraction temperatures may also improve efficiency because of enhanced removal rates, for example, by overcoming activation energy barriers to diffusion through micropores. Studies examining the mechanisms of matrix, modifier, and temperature effects in analytical SFE typically examine changes in extraction rates and efficiencies caused by varying extraction procedures. For example, comparisons have been made between extractions (i) conducted dynamically and for static/dynamic combinations of the same duration,8,19 (ii) from laboratory-spiked and field-contaminated samples,20,21 (iii) at different temperatures, pressures, or modifier concentrations,10,11 and (iv) at different flow rates.19 These studies seek to infer mechanisms from qualitative comparison of results obtained following different procedures or from multiparameter rate model fits. Models of the SFE process typically identify several potential rate-limiting steps including desorption from the active site, diffusion through an organic matrix or micropores to reach the particle surface, diffusion across a stagnant fluid boundary layer, and mass transport out of the extraction cell. Models have been developed that focus on a single step, lump several steps together, or treat each separately.22-30 Most studies have fit all model parameters simultaneously from a single experiment, creating uncertainty about the fundamental significance of the resulting values. Two exceptions to this generalization should be noted. In the first, a deconvolution approach was employed to mathematically separate desorption and mass transport effects by comparing removal of spiked (labeled) and “native” analytes from the same (15) Fahmy, T.; Paulaitis, M.; Johnson, D.; McNally, M. E. Anal. Chem. 1993, 65, 1462-1469. (16) Debendetti, P. G.; Kumar, S. K. AIChE J. 1988, 34, 645-657. (17) Kelley, F. D.; Chimowitz, E. H. AIChE J. 1990, 36, 1163-1175. (18) Miller, D. J.; Hawthorne, S. B.; Clifford, A. A.; Zhu, S. J. Chem. Eng. Data 1996, 41, 779-786. (19) Hawthorne, S. B.; Galy, A. B.; Schmitt, V. O.; Miller, D. J. Anal. Chem. 1995, 67, 2723-2732. (20) Burford, M. D.; Hawthorne, S. B.; Miller, D. J. Anal. Chem. 1993, 65, 14971505. (21) Langenfeld, J. J.; Hawthorne, S. B.; Miller, D. J.; Pawliszyn, J. Anal. Chem. 1995, 67, 1727-1736. (22) Recasens, F.; McCoy, B. J.; Smith, J. M. AIChE J. 1989, 35, 951-958. (23) Tan, C. S.; Liou, D. C. AIChE J. 1989, 35, 1029-1031. (24) Bartle, K. D.; Clifford, A. A.; Hawthorne, S. B.; Langenfeld, J. J.; Miller, D. J.; Robinson, R. J. Supercrit. Fluids 1990, 3, 143-149. (25) Bartle, K. D.; Boddington, T.; Clifford, A. A.; Hawthorne, S. B. J. Supercrit. Fluids 1992, 5, 207-212. (26) Clifford, A. A.; Burford, M. D.; Hawthorne, S. B.; Langenfeld, J. J.; Miller, D. J. J. Chem. Soc., Faraday Trans. 1995, 91, 1333-1338. (27) Akman, U.; Sunol, A. K. AIChE J. 1991, 37, 215-224. (28) Kothandaraman, S.; Ahlert, R. C.; Venkataramani, E. S.; Andrews, A. T. Environ. Prog. 1992, 11, 220-222. (29) Pawliszyn, J. J. Chromatogr. Sci. 1993, 31, 31-37. (30) Madras, G.; Thibaud, C.; Erkey, C.; Akgerman, A. AIChE J. 1994, 40, 777785.
sample in a single experiment.29 Although capable of fitting the experimental data, the model still employed several adjustable parameters which varied across experiments, suggesting a failure to accurately capture all fundamental processes. A second study fit SC CO2 desorption data from soils and activated carbon using a rate model with no adjustable parameters; all parameters were obtained by independent measurement or from literature correlations.30 The success of this model may be limited in analytical applications because it has only been applied to chemicals that were spiked from the supercritical phase during short contact times. The model is not expected to predict the well-established extraction rate differences between spiked and native analytes.20,21 Differences in analyte-matrix affinity have been suggested as a potential source of matrix effects.6 Models of analytical SFE, often derived from chromatographic models, typically employ a retention factor or partition coefficient to quantify chemicalsurface interactions.26,29 Unlike chromatographic packing materials, heterogeneous environmental samples feature a wide range of active site types. The energetic heterogeneity of natural surfaces toward nonpolar organic compounds and the resulting nonlinearity of sorption isotherms describing an analyte’s solidfluid distribution have been demonstrated previously for aqueous systems.31 Nonlinear isotherms are frequently fit by the Freundlich model. The form of the model and its relation to the partition coefficient (K) and the chromatographic retention factor (k) are given by
qe ) KFCen
(1)
K ) qe/Ce ) k(VF/VSF) ) KFCen-1
(2)
where qe and Ce are the solid and fluid (SC CO2) phase analyte concentrations, KF and n are the Freundlich parameters, VF and VS are the fluid and solid phase volumes in the extraction vessel, and F is the solid density. Nonlinear isotherms are potentially significant in SFE because desorption partition coefficients may become increasingly unfavorable as analyte is removed from progressively higher energy active sites. If n < 1, a common situation for single-layer adsorption on heterogeneous surfaces, the partition coefficient will increase as Ce decreases, and extraction rates will decline. Previous studies measuring equilibrium distributions of organic compounds between SC CO2 and natural sorbents have studied either a single soil or a single solid loading, and none has related the results to analytical extraction efficiencies.32-34 The objectives of the present study were to examine the effect of matrix characteristics and common extraction variables on equilibrium analyte distribution between environmental sorbents and SC CO2 and to relate these differences to analyte extractability. SFE efficiencies and rates for a model solute, phenanthrene, were measured from five well-characterized natural sorbents (four soils and a shale) as a function of methanol modifier content. Desorption equilibria of phenanthrene from the same sorbents were measured in SC CO2 at different temperatures, pressures, and (31) Young, T. M.; Weber, W. J., Jr. Environ. Sci. Technol. 1995, 28, 92-97. (32) Andrews, A. T.; Ahlert, R. C.; Kosson, D. S. Environ. Prog. 1990, 9, 204210. (33) Erkey, C.; Madras, G.; Orejuela, M.; Akgerman, A. Environ. Sci. Technol. 1993, 27, 1225-1231. (34) Hess, R. K.; Erkey, C.; Akgerman, A. J. Supercrit. Fluids 1991, 4, 47-52.
Analytical Chemistry, Vol. 69, No. 8, April 15, 1997
1613
Table 1. Summary of Sorbent Characteristics
organic matter (wt %) carbona (wt %) hydrogena (wt %) nitrogena (wt %) moistureb (wt %) surface areac (m2/g) d
Ohio shale
Chelsea
Webster
Houghton
Wurtsmith
5.68 2.44 0.65 ndd 0.90 13.7
10.98 5.60 0.60 0.29 2.85 3.92
7.24 2.97 0.50 0.14 3.36 9.46
79.00 46.10 4.08 3.10 14.49 0.97
0.13 0.021 ndd ndd 0.046 0.72
a Determined by high-temperature combustion. b Measured for unspiked soils prior to spiking and freeze-drying. c Determined by N BET method. 2 Value not statistically different from zero.
modifier concentrations to establish the effect of these variables on analyte-matrix affinity at varied solid-phase analyte concentrations. The resulting data provide a fuller understanding of how matrix characteristics influence the extraction process. EXPERIMENTAL SECTION Sorbents. Five natural materials were collected as model environmental matrices. Three surface soils (Webster, Chelsea, and Houghton) were selected, ranging in organic matter content from 3 to 80%. A sandy material from the unsaturated zone (Wurtsmith) was selected to represent low organic matter, predominantly mineral materials. A shale (Ohio) was included because it was previously identified as having an unusually high capacity for phenanthrene in aqueous systems.31 The collection and characterization of these samples, as well as other experimental procedures employed in this work, are described in detail elsewhere.35-37 All sorbents were air-dried and sieved to remove particles larger than 2 mm, and then the remaining material was crushed to pass a 250 µm sieve (#60 U.S. standard sieve size) prior to spiking. Total carbon, nitrogen, and hydrogen contents of the sorbents were measured by combustion at 1050 °C using an elemental analyzer (Leco Corp., St. Joseph, MI). Specific surface areas were determined by fitting low-temperature nitrogen adsorption data (Autosorb I, Quantachrome Corp., Boynton Beach, FL, or Micromeritics Instrument Corp., Norcross, GA) to the BET isotherm model to obtain monolayer capacities. Matrix characteristics are summarized in Table 1. Spiking Procedure. Spiking was accomplished by contacting an aqueous phenanthrene (98+%, Aldrich Chemical Co., Milwaukee, WI) solution with sorbent in batch reactors tumbled endover-end for 30 days. The procedure was designed to eliminate artificially high SFE efficiencies resulting from rapid spiking by solvent evaporation.20 The contact time chosen is many times longer than those employed in previous studies with spiked sorbents but is still far shorter than the times that many fieldcontaminated materials have been exposed to pollutants. For the sorbents employed here, little difference in aqueous desorption was observed when comparing sorbents exposed to phenanthrene for 30 days and 1 year.35 However, it has been demonstrated that the sorption of nonpolar organic chemicals may continue for more than a year and that desorption half-lives may be on the order of decades for long-contaminated materials.38,39 The spiking procedure thus was expected to approximate a short-term field (35) Young, T. M. Ph.D. Dissertation, University of Michigan, 1996. (36) Weber, W. J., Jr.; Young, T. M. Environ. Sci. Technol., in press. (37) Young, T. M.; Weber, W. J., Jr. Environ. Sci. Technol., in press. (38) Ball, W. P.; Roberts, P. V. Environ. Sci. Technol. 1991, 25, 1223-1237.
1614 Analytical Chemistry, Vol. 69, No. 8, April 15, 1997
Figure 1. Schematic of supercritical fluid extraction apparatus.
contamination scenario but may have resulted in an overestimate of phenanthrene extractability from long-contaminated materials. Following the contact period, sorbent slurries were filtered (Type GF/A, Whatman Inc., Clifton, NJ), freeze-dried for approximately 24 h, crushed, and sieved to pass a 250 µm sieve. Spiked sorbents were split into representative subsamples using a stainless steel riffle splitter (Soiltest Inc., Lake Bluff, IL) and stored in individual glass vials in a desiccator with CaSO4 desiccant until use. At least six different solid loading levels were prepared for each sorbent. The following ranges of loading levels were prepared: Ohio shale, 7.9-1336 µg/g; Chelsea soil, 0.64-228 µg/ g; Webster soil, 1.6-208 µg/g; Houghton soil, 18.6-1203 µg/g; and Wurtsmith soil, 0.20-7.3 µg/g. Maximum loading levels differed because of the different sorption capacities of these materials from aqueous solution, while lower limits were determined from analytical detection limit considerations. Supercritical Fluid Extraction. SFE experiments were conducted in a commercial system (Prepmaster, Accutrap and Modifier Pump, Suprex Corp., Pittsburgh, PA), shown in Figure 1, that was modified to allow closed recirculation of SC CO2. The commercial unit consisted of a microprocessor control module, a piston pump, a cosolvent addition pump, an oven, an extraction cell, a pair of two-position six-port valves (V1 and V2), a heated variable restrictor to control CO2 flow rates, a cryogenically cooled solute trap, a pump to recover solute by flushing the trap, and a fraction collector. Modifications allowed the system to be used to recirculate SC CO2 through the extraction cell in a closed loop to attain a desorptive equilibrium. Components added to the system included a magnetically coupled gear pump (Micropump Corp., Vancouver, WA), a variable-wavelength UV detector (critical extraction monitor, Milton Roy Corp., Rochester, NY), and a (39) Steinberg, S. M.; Pignatello, J. J.; Sawhney, B. L. Environ. Sci. Technol. 1987, 21, 1201-1208.
sample loop to collect SC CO2 samples for off-line analysis. The apparatus was similar to those used by others to measure solubility or adsorption.12,32 High-purity CO2 (SFC grade, Scott Specialty Gases, Plumsteadville, PA) was used for all supercritical fluid experiments. Contaminated soil samples were packed dry into the extraction vessel until it was full, and a consistent soil mass and vessel size were employed in all experiments with a particular soil. The masses and vessel sizes employed were as follows: Houghton soil, 0.45 g, 1 mL; Ohio shale, 0.75 g, 1 mL; Chelsea soil, 1.0 g, 1 mL; Webster soil, 1.0 g, 1 mL; and Wurtsmith soil, 4.5 g, 3 mL. Most experiments consisted of two steps, an equilibrium stage of approximately 100-400 min duration, followed by a dynamic extraction. In the equilibrium stage, the system was brought to the specified temperature and pressure, and recirculation was commenced using the magnetically coupled gear pump. Recirculation continued until a stable ultraviolet absorbance reading was obtained (1.5-2 h). A sample of SC CO2 was taken and the sample analyzed off-line using HPLC. The sequence was continued for up to four pressures for a given sample. Isotherms with methanol modifier were determined by a similar procedure with the addition of a known volume of methanol via the sample loop following the initiation of recirculation. A 60 min dynamic extraction (310 atm, 50 °C) was performed at the conclusion of each isotherm sequence with varying concentrations of methanol modifier (0, 1, or 5 mol %). Condensed phase CO2 flow rates measured at the pump were 2 mL/min in all experiments. The heated restrictor was maintained at 75 °C, and the cryogenic trap was cooled to 10 °C for experiments with no modifier and 20 °C for experiments with methanol modifier. The dynamic extraction was divided into six steps of 5, 5, 5, 10, 15, and 20 min, and the cryogenic trap was flushed with methanol after each step. Under these collection conditions, recovery of phenanthrene from aluminum foil boats placed in empty extraction vessels was 102% ((5.6% RSD) and was essentially complete in the first 5 min period. To complete the mass balance in soil experiments, the sample was subjected to a 24 h post-SFE Soxhlet extraction with methanol. Mass recoveries compared to initial Soxhlet-derived solid loadings averaged 90% and were between 60 and 120% for all experiments reported. Phenanthrene concentrations were quantified using HPLC (Model 1050, Hewlett-Packard Corp., Palo Alto, CA) with diode array and fluorescence detection (Model 1046A, Hewlett-Packard Corp.). An isocratic 80/20 acetonitrile/water carrier was employed with a C18 ODS reverse phase column (2 mm × 250 mm, Phenomenex Corp., Torrance, CA). RESULTS AND DISCUSSION The effects of temperature, pressure, and modifier on analyte desorption were explored by measuring the equilibrium distribution of phenanthrene between each of the five solids and SC CO2. In Figures 2 and 3, desorption isotherms for two of the materials, Webster soil and Ohio shale, are compared for six combinations of temperature (40 or 60 °C), pressure (120 or 310 atm), and modifier concentration (0 or ∼6-9 mol %). Isotherms were welldescribed by the Freundlich model (R2 > 0.98) and were generally nonlinear when no methanol was added (0.67 < n < 1.05), becoming approximately linear following methanol addition (0.91 < n < 1.13). Capacity factors varied widely with conditions and between sorbents (6.5 < KF < 204). Freundlich model fits are shown as lines in Figures 2 and 3.
Figure 2. Effect of temperature, pressure, and modifier addition on phenanthrene desorption isotherms from Ohio shale in SC CO2.
Figure 3. Effect of temperature, pressure, and modifier addition on phenanthrene desorption isotherms from Webster soil in SC CO2.
The effect of each variable on desorption was qualitatively consistent with its effect on phenanthrene solubility in SC CO2: sorption was reduced at higher pressures and modifier concentrations, and the direction of the temperature effect depended on whether the pressure was within (120 atm) or outside (310 atm) the retrograde region. The magnitude of each variable’s impact differed between the two matrices, however. The pressure increase from 120 to 310 atm resulted in a 15-fold reduction in sorption capacity (ratio of KF values) for Ohio shale at 60 °C but only a 4-fold reduction for Webster soil. For comparison, the corresponding increase in phenanthrene solubility between 120 and 310 atm is 22-fold. Methanol modifier addition at 120 atm and 40 °C proved far more effective in stimulating desorption from Webster soil (9-fold reduction in KF) than from the Ohio shale (4-fold reduction). The impact of temperature, pressure, and methanol modifier on Freundlich capacity factors for each of the sorbents is summarized in Table 2. A complete listing of Freundlich isotherm parameters is reported elsewhere.36,37 Identifying the source of these matrix-specific responses to SFE conditions and their impact on efficiencies and rates was a major goal of this study. Results of 60 min dynamic extractions at 310 atm and 50 °C for the five matrices at several methanol modifier concentrations are reported in Table 3. Extraction efficiencies were calculated relative to 24 h Soxhlet extraction with methanol. Phenanthrene extraction efficiencies ranged from 22 to 85% when no methanol Analytical Chemistry, Vol. 69, No. 8, April 15, 1997
1615
Table 2. Effect of Extraction Variables on Isotherm Parameters for Different Matrices sorbent Chelsea Ohio shale Webster Houghton Wurtsmith
temperature responsivenessa
pressure responsivenessb
methanol responsivenessc
predicted time to 90% removald
2.22 2.83 1.13
1.72 5.42 1.75 1.43 1.42
3.71 2.78 11.3 5.25 2.53
11.4 3.79 39.3 28.6 3.92
a K (40 °C)/K (60 °C) at 310 atm. b K (120 atm)/K (310 atm) at 50 °C. c K (0% methanol)/K (7.4 mol % methanol) at 120 atm and 50 °C. d NLE F F F F F F model prediction of time (in minutes) required to extract 90% of phenanthrene from a 40 µg/g sample.
Table 3. Effect of Sorbent Type and Modifier Concentration on Supercritical Fluid Extraction Efficiency at 310 atm and 50 °C 0 mol % methanol
1 mol % methanol
5 mol % methanol
sorbent
efficiency (%) (%RSD)
obs
efficiency (%) (%RSD)
obs
efficiency (%) (%RSD)
obs
Chelsea Ohio shale Webster Houghton Wurtsmith
36.5 (11.0) 84.8 (11.0) 24.5 (15.5) 22.0 (14.1) 74.8 (4.7)
7 9 15 5 5
48.0 (21.0) 92.9 (17.0)
10 11
66.0 (17.3) 92.4 (24.5) 70.3 (7.2) 78.9 (5.2) 109.8 (17.9)
12 5 14 6 3
Figure 4. Cumulative extraction of phenanthrene from five natural sorbents during dynamic extraction at 310 atm and 50 °C.
Figure 5. Comparison of phenanthrene extraction rates from five natural sorbents at 310 atm, 50 °C, and 5 mol % methanol modifier.
was added, while the presence of modifier increased efficiencies to between 48 and 109%. Relative standard deviations for these experiments were somewhat high, at least in part because efficiency values were averaged over materials spiked at several different solid loadings. The values thus neglect extraction differences arising from isotherm nonlinearity (see below) and incorporate any uncertainty in the independent Soxhlet determinations of the initial loading levels. No attempt was made in this work to conduct quantitative supercritical fluid extractions. The extraction conditions without methanol modifier were selected to match the most stringent conditions for which isotherms were available for all sorbents, and those with methanol were selected to be the same to permit comparison. The amount of phenanthrene extracted during each time step of the 60 min extraction was summed to generate a cumulative desorption profile; results for dynamic extractions with and without methanol for each of the five sorbents are compared in Figures 4 and 5. All extractions rapidly reached a plateau within the first 10 min, beyond which little additional extraction occurred. However, significant differences in ultimate extraction efficiency were observed, ranging from 22% for the Houghton muck with no methanol to 109% for the Wurtsmith soil with 5 mol % methanol.
Error bars in the figures represent standard deviations of the cumulative extraction measured at each time point. As will be shown below, an important component of this variability arises from differences in initial solid-phase loadings. Errors were typically smaller when only a single solid loading was considered for each matrix. The cumulative extraction curves presented here differ from similar curves reported by others because an equilibrium phenanthrene distribution was established prior to commencing the dynamic step. In all cases, the amount of phenanthrene present in the fluid phase at the beginning of the dynamic step was less than that extracted in the first 5 min. Similar dynamic experiments conducted without an initial isotherm step showed no significant differences in cumulative extraction curves for any of the sorbents. The shapes of the extraction profiles are consistent with effluent concentration profiles and cumulative extraction curves obtained by other investigators.19-30 Ultimate phenanthrene removals for the sorbents varied in the order Ohio shale > Wurtsmith > Chelsea > Webster ≈ Houghton, were generally opposite to the isotherm capacities at 310 atm and 50 °C, which were Houghton > Webster > Chelsea > Ohio shale > Wurtsmith. Only the two lowest capacity materials,
1616 Analytical Chemistry, Vol. 69, No. 8, April 15, 1997
Figure 6. Nonlinear local equilibrium model results for dynamic SFE of phenanthrene from Ohio shale.
Figure 7. Nonlinear local equilibrium model results for dynamic SFE of phenanthrene from Webster soil.
Wurtsmith and Ohio shale, had extraction efficiencies that were reversed from their respective Freundlich capacity factors. A simple nonlinear local equilibrium (NLE) model was applied to the data to examine whether equilibrium analyte distributions under supercritical conditions could be used to predict SFE rate curves.40 The model assumes ideal plug-flow behavior (i.e., no longitudinal dispersion), no diffusional or mass transfer resistances, and instantaneous sorption reactions throughout the extraction vessel. The time required to reach a particular solid or fluid phase analyte concentration may be calculated as
t)
L 1 - dq 1+ F vx dC
[ (
) ]
where L is the extraction cell length, vx is the interstitial fluid velocity, is the void volume of the bed, and dq/dC is the isotherm slope at the desired concentration. This model effectively predicted desorption of organic chemicals from activated carbon in SC CO2 but failed to fit similar data for soil.33,41 Figures 6 and 7 show the NLE model fits to extraction rate data for Ohio shale and Webster soil at two different solid loadings. In each case, the model significantly overestimates ultimate solute removal, as might be expected for a model that neglects mass transfer resistances, predicting quantitative recoveries from each matrix. Despite its shortcomings, the model illustrates several important points about matrix effects. First, it provides a qualitative indication that extraction from Webster soil will be far more (40) Ruthven, D. M. Principles of Adsorption and Adsorption Processes; John Wiley & Sons: New York, 1984. (41) Madras, G.; Erkey, C.; Akgerman, A. Ind. Eng. Chem. Res. 1993, 32, 11631168.
Figure 8. Sensitivity of nonlinear local equilibrium model results for Webster soil to variations in the Freundlich n value.
difficult than that from Ohio shale under the selected conditions. Even in the absence of mass transfer limitations, almost 40 min would be required to attain 90% extraction efficiency from Webster soil contaminated with 40 µg/g phenanthrene, while only 3.8 min would be required for similarly contaminated Ohio shale (Table 2). The model also correctly indicates that a longer extraction will be required to achieve similar fractional removals for less contaminated samples of Webster soil. This effect results from the nonlinearity of the Webster isotherm. The sensitivity of cumulative extraction curves to isotherm nonlinearity is illustrated in Figure 8, which shows the dramatic impact that small changes in the Freundlich n value can have in the absence of mass transfer limitations. Rate models assuming constant capacity factors would necessarily attribute these differences to mass transfer effects. The large discrepancy between the NLE model and the extraction rate data clearly illustrates that rate limitations must be considered in optimizing SFE for heterogeneous environmental samples. The effectiveness of the model in qualitatively predicting extraction success using only equilibrium parameters suggests that the relevant rate parameters are closely related to measures of analyte-matrix affinity. Assuming that the analyte must diffuse past a number of active sites to be removed from the solid particle, it is expected that higher sorption affinity will correspond to lower effective diffusivity. Alternatively, the difference between the extraction rate plateau and the complete removal predicted by the NLE model may relate to the desorption-resistant fraction of sorbed contaminants previously noted in the environmental science literature.39 Phenanthrene tended to be more completely extracted from low organic matter, high surface area materials at the selected supercritical conditions. At 310 atm and 50 °C, extraction efficiencies exhibited a weak negative correlation with matrix organic matter content (F ) -0.57) and a weak positive correlation with matrix-specific surface area (F ) 0.33). The role of organic matter in increasing sorption and decreasing extractability has been noted previously.34,42 It is likely that differences in natural organic matter (NOM) structure, particularly its rigidity and polarity (e.g., O/C ratio), account for the variation in phenanthrene extractability not explained by organic matter quantity. Geologically “young” NOM present in surface soils tends to be more flexible and contain higher densities of polar functional groups than the diagenetically altered NOM present in ancient kerogen (42) Rochette, E. A.; Harsh, J. B.; Hill, H. H., Jr. Environ. Sci. Technol. 1996, 30, 1220-1226.
Analytical Chemistry, Vol. 69, No. 8, April 15, 1997
1617
materials.43 These differences have been shown to have important implications for the sorption of organic chemicals from aqueous solutions.31 Analogies between NOM and synthetic polymers have proven useful in understanding sorption of organic chemicals from aqueous solution, and similar reasoning has been applied to desorption equilibria in SC CO2.36,37 The effectiveness of elevated pressures may differ among sorbents because the amorphous NOM in surface soils swells at high CO2 activities, while the more rigid kerogen in shale swells far less. Since sorption in polymers is appropriately expressed on a volume fraction basis, the expanded phase volume accompanying swelling increases the sorption capacity of the surface soils, counteracting the effect of increased solubility. Studies with synthetic polymers reveal similar differences in sorption and swelling between amorphous and glassy materials in SC CO2.44,45 Solubility increases obtained through temperature increases do not result in similar swelling and, therefore, are more effective in stimulating desorption. Strong analyte-matrix interactions have been invoked to explain the matrix effect. To test this hypothesis, enthalpies of phenanthrene sorption from Webster and Chelsea soils and Ohio shale were calculated from the temperature dependence of isotherm parameters.37 Desorption enthalpies from an ideal gas reference state ranged from -106.5 to -70.4 kJ/mol and were not statistically distinguishable from phenanthrene’s heat of condensation (-∆HVAP) of -70.7 kJ/mol.46 The results, similar to vapor phase sorption results for dry soils, confirm the presence of relatively high-energy physical sorption interactions with surface functional groups in natural organic matter.47 Such interactions might have occurred, for instance, between acidic organic matter sites in the freeze-dried soils and polarizable phenanthrene molecules. These binding interactions appear to be disrupted at higher extraction temperatures. Following the 60 min dynamic extraction for Chelsea soil at 310 atm and 50 °C with a second 60 min extraction at 310 atm and 100 °C caused an immediate jump in the cumulative extraction curve, reaching a second plateau within 10 min.35 Similar data have been reported by others.26 Although no phenanthrene solubility data in SC CO2 are available at these high temperatures, solubility increases by a factor of about 2.6 when system temperature is raised from 50 to 100 °C according to calculations using the Peng-Robinson equation of state. Measured solubilities of other polycyclic aromatic hydrocarbons show similar temperature dependence.18 Far larger solubility increases achieved by varying pressure at constant temperature had little effect on phenanthrene desorption isotherms. These observations support previous hypotheses that increases in extraction efficiency with temperature result from disruption of analytematrix binding interactions.6 Addition of methanol modifier at 120 atm and 50 °C dramatically reduced the affinity of all sorbents for phenanthrene, and the isotherms became nearly linear. Capacities remained closely correlated with organic matter content, but differences among (43) Hatcher, P. G.; Breger, I. A.; Maciel, G. E.; Szeverenyi, N. M. In Humic Substances in Soil Sediment and Water; Aiken, G. R., McKnight, D. M., Wershaw, R. L., MacCarthy, P., Eds.; John Wiley & Sons: New York, 1985. (44) Shim, J.; Johnston, K. P. AIChE J. 1989, 35, 1097-1106. (45) Wissinger, R. G.; Paulaitis, M. E. Ind. Eng. Chem. Res. 1991, 30, 842-851. (46) Johnston, K. P.; Ziger, D. H.; Eckert, C. A. Ind. Eng. Chem. Fundam. 1982, 21, 191-197. (47) Rao, P. S. C.; Ogwada, R. A.; Rhue, R. D. Chemosphere 1989, 18, 21772191.
1618
Analytical Chemistry, Vol. 69, No. 8, April 15, 1997
sorbents were far smaller. The reduction in sorption far exceeded the measured solubility increase of 21%.37 Methanol modifier appears to act by competitively sorbing to high-energy polar organic matter sites, displacing polarizable phenanthrene molecules and decreasing the heterogeneity of available sites. Unsubstituted aromatics have previously been proven ineffective in competing with methanol for stationary phase active sites in supercritical chromatography.48 The largest impact of methanol addition on extraction efficiencies was observed for the Webster and Houghton soils, for which recoveries increased by factors of 2.9 and 3.6, respectively, upon addition of 5 mol % methanol compared to extractions conducted without cosolvent. These sorbents also showed the largest percentage reduction in isotherm capacity at 120 atm when methanol was added at similar concentrations (Table 2). Isotherms and extraction efficiencies for Ohio shale were affected far less by the presence of methanol. These observations are consistent with enhanced methanol sorption on the more polar organic matter present in dry surface soils (Houghton and Webster) relative to kerogens (Ohio shale). Addition of 1 mol % methanol produced nearly quantitative recoveries for the Ohio shale and increased recovery for the Chelsea soil by 33%. Increasing the modifier content to 5 mol % raised the recovery for Chelsea soil an additional 38%. The similar impact of the initial small methanol addition (from 0 to 1 mol %) and the second larger increment (from 1 to 5 mol %) are consistent with a competitive displacement mechanism but do not support a solubility enhancement mechanism. Cumulative extraction curves for Chelsea soil at 0, 1, and 5 mol % methanol all had similar shapes, reaching a plateau within the first 10 min of the extraction.35 Thus, rate enhancement resulting from clay swelling by polar modifiers may not be an important mechanism of modifier interaction for removing nonpolar analytes as it was for urea herbicides.15 The present results do support the notion that matrix-cosolvent interactions are an important mechanism for polar modifier effects.7 CONCLUSIONS Combining equilibrium and rate results clearly indicates that differences in extraction efficiency across matrix types relate primarily to the strength and nature of analyte-matrix interactions. Isotherm and enthalpy data lend further support to previous observations that proper modifier selection and increased system temperature are successful extraction strategies because they disrupt physical analyte-matrix bonding. The primary reason for the ineffectiveness of increased pressure for extracting organic solutes from certain natural organic matrices may relate to sorbent swelling upon exposure to high-pressure CO2. Increasing temperature at constant pressure tends to decrease matrix swelling by decreasing the activity of the solvent phase (reduced fluid density) and thereby serves to increase solute removal by three mechanisms: reducing swelling-induced sorption, improving solubility, and overcoming exothermic solute-surface interactions. A nonlinear local equilibrium model was used to illustrate the potentially significant impact of isotherm nonlinearity and capacity differences between sorbents on extraction rates and efficiencies. The presence of a limited number of high-energy binding sites on heterogeneous environmental samples, evidenced in this study by high desorption enthalpies and nonlinear sorption isotherms, can significantly impair SFE unless appropriate measures (in(48) Lochmu ¨ ller, C. H.; Mink, L. P. J. Chromatogr. 1989, 471, 357-366.
creased temperature and/or modifier addition) are taken to lower energetic barriers to desorption. ACKNOWLEDGMENT This research was funded in part by the U.S. Environmental Protection Agency, Office of Research and Development, Risk Reduction Engineering Laboratory, Cincinnati, OH, under Cooperative Agreement CR 818213-01-0. The project officer was Dr. James Ryan. The authors thank Kevin Couch, an undergraduate
research assistant at the University of California, Davis, for his help in conducting the experimental work.
Received for review October 3, 1996. Accepted January 23, 1997.X AC961014J X
Abstract published in Advance ACS Abstracts, March 1, 1997.
Analytical Chemistry, Vol. 69, No. 8, April 15, 1997
1619