Partitioning of Organic Chemicals to Polyacrylate ... - ACS Publications

Wouter H. J. Vaes,* Casper Hamwijk, En˜ aut Urrestarazu Ramos, Henk J. M. Verhaar, and. Joop L. M. Hermens. Research Institute of Toxicology (RITOX),...
0 downloads 0 Views 151KB Size
Anal. Chem. 1996, 68, 4458-4462

Partitioning of Organic Chemicals to Polyacrylate-Coated Solid Phase Microextraction Fibers: Kinetic Behavior and Quantitative Structure-Property Relationships Wouter H. J. Vaes,* Casper Hamwijk, En˜aut Urrestarazu Ramos, Henk J. M. Verhaar, and Joop L. M. Hermens

Research Institute of Toxicology (RITOX), Utrecht University, P.O. Box 80176, NL-3508 TD, Utrecht, The Netherlands

The solid phase microextraction technique uses polymercoated fused-silica fibers to extract organic chemicals from an aqueous or gaseous phase. In the current paper, the partitioning behavior of organic chemicals from water to polyacrylate coating is described in terms of a twocompartment model and first-order kinetics. Experimental results show that, through agitation, it is possible to achieve a sufficiently small aqueous diffusion layer around the fiber to prevent aqueous diffusion from being the limiting factor in the absorption process. This then implies that the equilibration time is completely determined by the polyacrylate phase. In addition, the kinetic rate constants derived from the experimental results, viz. the uptake rate constant and polyacrylate-water partition coefficient, are modeled by multivariate techniques, using physicochemical and quantum chemical descriptors. These models clearly show that, besides hydrophobicity expressed as the octanol-water partition coefficient, the energy of the lowest unoccupied molecular orbital (and related properties) and the most positive charge on any hydrogen atom in the molecule are important descriptors. This indicates that hydrogen bonding plays a significant role in polyacrylate-water partitioning. The models that are presented can be used to predict absorption profiles of organic chemicals to polyacrylate-coated fibers, thereby giving the opportunity to predict the kinetics (including equilibration times) by computations alone. Solid phase microextraction (SPME) is an extraction technique that uses polymer-coated fibers to extract compounds from an aqueous1,2 or gaseous3 phase. After extraction, the fiber is transferred directly to the injector of a gas chromatograph in which the compounds are thermally desorbed and subsequently analyzed and quantified.4,5 The absorption of compounds from the sample to the fiber is an equilibrium partitioning process that has been described in great detail by Louch et al.6 The time after which equilibrium will be reached depends on the compound, fiber coating material, * E-mail: [email protected]. (1) Buchholz, K. D.; Pawliszyn, J. Environ. Sci. Technol. 1993, 27, 2844. (2) Potter, D. W.; Pawliszyn, J. J. Chromatogr. 1992, 625, 247. (3) Zhang, Z.; Pawliszyn, J. Anal. Chem. 1993, 65, 1843. (4) Arthur, C. L.; Pawliszyn, J. Anal. Chem. 1990, 62, 2145. (5) Boyd-Boyland, A. A.; Chai, M.; Luo, Y. Z.; Zhang, Z.; Yang, M. J.; Pawliszyn, J. B.; Go´recki, T. Environ. Sci. Technol. 1994, 28, 569A. (6) Louch, L.; Motlagh, S.; Pawliszyn, J. Anal. Chem. 1992, 64, 1187.

4458 Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

thickness of the coating, and extent of mixing of the sample solution.6 The equilibration time can be measured by exposing the fiber to standard solutions of a chemical during increasing time intervals, until a steady state is reached.4 In this study, the partitioning process of 11 polar and 8 nonpolar narcosis compounds,7 which are not dissociated at the experimental conditions, to polyacrylate-coated fibers is investigated. The aim of the study is to develop quantitative structureproperty relationships (QSPR), using physicochemical and quantum chemical descriptors, to predict the kinetics of organic chemical partition behavior from water to polyacrylate-coated fibers. Based on the QSPR, KSPME for structurally related compounds can be predicted. In addition, the study may lead to more insight into the interactions between the coating of the fiber and the test chemicals. In order to have a broad variety in chemical properties (hydrophobic as well as electronic), the test chemicals were selected by statistical design. This design for the 11 polar chemicals is described by Urrestarazu et al.8 The selection of the eight nonpolar chemicals was based on their hydrophobicity expressed as the octanol-water partition coefficient (log Kow). THEORETICAL SECTION The kinetics of partitioning to SPME fibers have been described by Louch et al.6 They derived an expression to describe the absorption of organic compounds in a perfectly agitated solution. In this model, the rate of diffusion within the coating on the fiber is assumed to limit the rate of uptake by the fiber. However, on the basis of their experimental results, they concluded that the extraction process in an agitated solution is controlled by diffusion of the analyte through the thin stagnant aqueous layer located around the fiber. This observation led to the conclusion that diffusion within the fiber can be neglected when the absorption process is modeled. In the current study, we consider both the water phase and the polyacrylate coating as two separate compartments. Assuming first-order kinetics, the uptake process can then be described by the following expression:

d[X]f,t ) k1[X]a,t - k2[X]f,t dt

(1)

in which k1 is the uptake rate constant (from water to fiber), k2 (7) Verhaar, H. J. M.; van Leeuwen, C. J.; Hermens, J. L. M. Chemosphere 1992, 25, 471. (8) Urrestarazu Ramos, E.; Vaes, W. H. J.; Verhaar, H. J. M.; Hermens, J. L. M., submitted to Environ. Sci. Pollut. Res. S0003-2700(96)00747-0 CCC: $12.00

© 1996 American Chemical Society

the elimination rate constant (from fiber to water), and [X]f,t and [X]a,t are the concentration of a compound X in the fiber coating and the aqueous phase at time t, respectively. Assuming an infinite sample volume, eq 2 is valid where [X]a,t)0 is the aqueous

[X]a,t ) [X]a,t)0

(2)

concentration at time zero. Substitution of eq 2 into eq 1 and subsequent integration yields eq 3.

[X]f,t ) [X]a,t)0(k1/k2)(1 - e-k2t)

(3)

This equation is only valid if the extraction efficiency is very small (i.e., [X]a,t ) [X]a,t)0) for all sampling times. This equation has been used successfully before to describe the kinetics of organic compounds to SPME fibers.9 Extrapolating eq 3 to infinite time shows that

[X]f,t [X]a,t)0

)

k1 ) KSPME k2

(4)

Therefore, a nonlinear curve-fitting of eq 3 to an experimental absorption profile can be used to calculate the polyacrylate-water partition coefficient. EXPERIMENTAL SECTION Apparatus and Reagents. SPME devices with 1 cm fibers coated with 85 µm polyacrylate film were purchased from Supelco (Bellefonte, PA). These devices were employed for experiments with aniline, 1-butanol, 3-pentanol, 1-hexanol (Merck, Darmstadt, Germany), nitrobenzene, 3-nitroaniline, chlorobenzene (Fluka AG, Buchs, Germany), and 2-butoxyethanol (Aldrich, Steinheim, Germany). In order to accomplish negligible depletion in the samples, to comply with the need to keep eq 3 valid, the fibers were cut to a length of 1 mm for use in experiments with 2-allylphenol, 4-chloro-3-methylphenol, quinoline, 1,3,5-trichlorobenzene (Aldrich), 4-n-pentylphenol (Pfaltz & Bauer, Waterbury, CT), N,N-dimethylaniline, 2-phenylphenol (Sigma, St. Louis, MO), 2,4,5-trichlorotoluene (Janssen Chimica, Geel, Belgium), 2-nitrotoluene, p-xylene (Fluka AG), and 2,4,5-trichloroaniline (Riedelde Hae¨n AG, Seelze, Germany). Fibers were conditioned for 1 h at 300 °C in a helium atmosphere in the split injector of a GC to clean the fibers. All analyses were performed using a Carlo Erba MFC500 gas chromatograph (Carlo Erba Instruments, Milan, Italy) equipped with a split/splitless injector controlled by a SL-516 module. Either a 30 m × 0.32 mm fused-silica DB 5.625 column (J&W Scientific, Folson, CA) with a 0.25 µm film thickness or a 30 m × 0.32 mm DB-WAX (J&W Scientific) with a 0.5 µm film thickness was used. The detector used was an electron capture detector, a flame ionization detector, or a QMD 1000 mass spectrometer (Carlo Erba Instruments, electron impact energy 70 eV, detector multiplier at 400 V). The GC injector was kept at 275 °C. The injector was used in the splitless mode for 5 min during which the injection was substantiated. Subsequently, it was put in the split mode for the rest of the run. Directly after returning to the (9) Vaes, W. H. J.; Urrestarazu Ramos, E.; Verhaar, H. J. M.; Seinen, W.; Hermens, J. L. M. Anal. Chem. 1996, 68, 4463 (following paper in this issue).

split mode, the fiber was released from the injector. The oven was kept at 50 °C during the 5 min injection period in order to focus the sample analytes, after which the temperature was raised at 40 °C/min to 250 °C. For all compounds, carry-over due to incomplete desorption was determined by injecting the same fiber for a second time. Procedure: Absorption Profiles and Rate Constant Determinations. Several mixtures of the compounds were dissolved in a 0.10 M potassium phosphate (Merck) buffer solution (pH 7.4) in a volumetric flask. These solutions were stirred for 24 h to achieve complete dissolution. From each solution 1.5 mL was quickly transferred to 2.0 mL sample vials (PhaseSep, Waddinxveen, The Netherlands) that contained a 2 mm × 5 mm spin bar. For each measurement, a new vial was prepared and all measurements were made in duplicate while the contents were stirred at 2000 rpm. The amount that was thermally desorbed from the fibers during injection was calibrated using standards in hexane (J.T. Baker, Deventer, The Netherlands) that were analyzed using the same experimental conditions as for the fibers. The concentrations in the coating of the fiber were calculated by assuming a fiber volume of 0.521 and 0.0521 µL for the 1 cm and 1 mm fibers, respectively (outer radius 140 µm, inner radius 55 µm). Equation 3 was fitted through the averages of each time interval using the NONLIN module of the Systat v. 5.0 program.10 Statistical calculations were performed on a Power Macintosh 6100/66. Reported correlation coefficients between measured and fitted values were corrected for the degrees of freedom. Calculation/Retrieval of Descriptors. Quantum chemical calculations were performed using the quantum chemical package Spartan11 running on an IBM RS/6000 workstation. All calculations were performed using the semiempirical AM1 Hamiltonian.12 Structures were entered directly in Spartan. Minimum energy structures were generated by preoptimizing all possible conformers using the built-in force-field optimizer. Lowest energy conformers were manually picked and subsequently submitted to a quantum chemical optimization run. Descriptors, HOMO (energy level of the highest occupied molecular orbital), LUMO (energy level of the lowest unoccupied molecular orbital), electronegativity (EN, calculated as 1/2(HOMO + LUMO)), hardness (η, calculated as 1/2(HOMO - LUMO)), dipole moment (µ), polarizability (RE), molecular surface area (MSA), and molecular volume (MV) were taken from or calculated from the self-consistent field wave function for the energy-minimized molecular structure. Q- (most negative charge on any non-hydrogen atom) and Q+ (most positive charge on any hydrogen atom) were taken from charge lists calculated by fitting an electrostatic potential to a discrete localized charge distribution. A more detailed description is given elsewhere.13,14 Values of the octanol-water partition coefficient (log Kow) were retrieved from the MedChem Thor Database or calculated using the MedChem CLOGP3 algorithm.15 (10) Systat Inc., Evanston, IL, 1989. (11) Hehre, W. J.; Burke, L. D.; Shusterman, A. J. Spartan User’s Guide; Wavefunction Inc., Irvine, CA, 1993. (12) Dewar, M. J. S.; Zoebisch, E. G.; Healy, E. F.; Stewart, J. J. P. J. Am. Chem. Soc. 1985, 107, 3902. (13) Vehaar, H. J. M.; Urrestarazu Ramos, E.; Hermens, J. L. M. J. Chemom. 1996, 10, 149. (14) Cramer, C. J.; Famini, G. R.; Lowrey, A. H. Acc. Chem. Res. 1993, 26, 599. (15) Leo, D.; Weininger, D. MedChem Software Manual v. Software. Daylight Information Systems, Inc., Irvine, CA, 1989.

Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

4459

Table 1. Kinetic Constants for the Partition Process of 19 Organic Compounds to Polyacrylate-Coated Fibers and the Adjusted r2 of the Fitted (Nonlinear) Model

Figure 1. Examples of absorption profiles for chlorobenzene (2), 3-pentanol (O), and 1,3,5-trichlorobenzene (0). The lines represent the fitted models according to eq 3.

Data Analyses. All data analyses were performed on a Tulip Compact 486dx2/50 personal computer using the SCAN 1.1 package (Software for Chemometric Analysis, Minitab Inc., State College, PA). Rate constants k1 and k2 and partition coefficient KSPME were log transformed and analyzed using two approaches. First, a one-block partial least-squares projection to latent structures (PLS) analysis was used to separately model the rate constant (k1) and the partition coefficient (KSPME) of the chemicals from their calculated variables. The significant number of latent variables was determined by leave-one-out cross-validation.16 The dimension with lowest prediction error sum of squares (PRESS) was chosen, up to a maximum of four latent variables, according to Geladi and Kowalski.16 The predictive power of the model was calculated using the cross-validated correlation coefficient Q2, calculated as 1-PRESS/SSY.17 Second, a theoretical linear solvation energy relationship (TLSER) was developed using multiple linear regression (MLR) with log Kow as parameter describing hydrophobicity, HOMO and LUMO as covalent, and Q- and Q+ as ionic contributions to hydrogen bonding, using the three most important variables. The TLSER approach has been developed and applied by Cramer et al.14 RESULTS AND DISCUSSION Absorption Profiles and Rate Constant Determinations. Carry-over after injections was generally not observed and always below 0.1%. The depletion of the samples was always below 5%. Depletion was calculated for the longest sampling time tmax as 100% × (k1Vf/k2Va)(1 - e-k2tmax), in which Vf and Va represent the volumes of the fiber and the aqueous phase, respectively (data not shown). The one-compartment model that was used here to model the transfer of organic chemicals to the SPME fibers appears to be a good approximation of the absorption process. This is demonstrated in Figure 1. The correlation coefficients between measured and fitted values are all between 0.86 and 0.99. Values for log k1, log k2, and log KSPME are presented in Table 1. The uptake rate constant log k1 has a high variance (1.01) while this value is only 0.06 for the elimination rate constant, log k2. (16) Geladi, P.; Kowalski, B. R. Anal. Chim. Acta 1986, 185, 1. (17) Wold, S. Quant. Struct.-Act. Relat. 1991, 10, 191.

4460 Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

compound

log Kow

log k1a (min-1)

log k2b (min-1)

log KSPMEc

r2

2-butoxyethanol 1-butanol aniline 3-pentanol 2-nitroaniline nitrobenzene 1-hexanol quinoline 2-nitrotoluene N,N-dimethylaniline 2-allylphenol chlorobenzene 2-phenylphenol 4-chloro-3-methylphenol p-xylene 2,4,5-trichloroaniline 1,3,5-trichlorobenzene 4-n-pentylphenol 2,4,5-trichlorotoluene

0.83 0.88 0.90 1.21 1.37 1.85 2.03 2.03 2.30 2.31 2.64 2.90 3.09 3.10 3.15 3.69 4.19 4.24 4.78

-0.92 -0.68 0.02 -0.82 0.73 1.09 -0.08 0.59 1.51 1.39 1.61 0.87 1.85 1.49 1.28 1.89 2.12 1.90 2.15

-0.92 -0.72 -0.93 -0.95 -1.16 -0.76 -0.96 -0.85 -1.04 -0.74 -1.13 -1.09 -1.52 -1.28 -1.00 -1.65 -0.96 -1.28 -1.05

0 0.05 0.95 0.13 1.90 1.85 0.88 1.44 2.55 2.14 2.74 1.96 3.37 2.76 2.28 3.55 3.08 3.17 3.19

0.90 0.93 0.99 0.95 0.98 0.97 0.96 0.86 0.96 0.90 0.98 0.99 0.99 0.98 0.97 0.97 0.94 0.97 0.96

a Rate constant from aqueous to polyacrylate phase. b Rate constant from polyacrylate to aqueous phase. c Partition coefficient polyacrylatewater calculated as k1/k2.

Figure 2. Theoretical model, describing the kinetics of partitioning from water to a hydrophobic phase. Log k1 (rate constant from aqueous to hydrophobic phase), log k2 (rate constant from hydrophobic to aqueous phase), and log K (partition coefficient) are plotted vs the hydrophobicity. In the first part, the resistance in the hydrophobic phase is the rate-limiting step for accumulation in this phase, in the second part the resistance in the aqueous phase is rate limiting. (Modified from Gobas et al.21)

Consequently, the variance in partition coefficient log KSPME (1.32) depends mostly on k1. The influence of hydrophobicity on the kinetics of partitioning is discussed in several publications.18-21 A theoretical model, describing the rate-limiting processes in partitioning from water to a hydrophobic solvent is given in Figure 2. For chemicals with low hydrophobicity, the diffusive resistance in the hydrophobic phase is limiting the uptake process. Thus an increase in hydrophobicity results in an increasing k1 and a merely constant (18) Flynn, G. L.; Yalkowsky, S. H. J. Pharm. Sci. 1972, 61, 839. (19) Yalkowsky, S. H.; Slunick, T. G.; Flynn, G. L. J. Pharm. Sci. 1974, 63, 691. (20) Gobas, F. A. P. C.; Opperhuizen, A.; Hutzinger, O. Environ. Toxicol. Chem. 1986, 5, 637. (21) Gobas, F. A. P. C.; MacKay, D. Environ. Toxicol. Chem. 1987, 6, 495.

Table 2. Importance and Pseudoregression Coefficients (PRC) for PLS Analyses on log KSPME and log k1 log KSPME

intercept log Kow HOMO LUMO EN η µ RE MSA MV QQ+ Figure 3. Rate constants log k1 ([) and log k2 (O) as a function of the equilibrium constant (log KSPME) for the polyacrylate-coated fiber.

k2. Increasing the hydrophobicity further will result in a certain point where the uptake rate becomes so high that the aqueous layer cannot transport enough chemical to the hydrophobic phase to let k1 increase further. Therefore, k1 does not change anymore, due to limitation by diffusive resistance in the aqueous phase. From this point, k2 starts to decrease.18 The results of this study show that log KSPME is related to log k1 (r2 ) 0.96) and in minor extent to log k2 (r2 ) 0.42). In Figure 3, log k1 and log k2 are plotted vs their respective log KSPME. Log k1 increases with increasing log KSPME, while log k2 is relative constant. This strongly suggests that the internal resistance in the fiber coating is rate limiting for the partition process in an agitated solution, making a faster absorption than observed impossible. However, the kinetic behavior and rate-limiting processes may be completely different at lower stirring rates. It has been shown that the equilibration times strongly increase at low stirring rates,6,22 indicating that, under those conditions, the resistance in the aqueous phase is rate limiting. The observations from this study are not in accordance with the conclusion of Louch et al.6 that the extraction process is controlled by diffusion of the analyte through the thin stagnant aqueous layer located around the fiber. It should be pointed out that Louch et al.6 performed their calculations based on a diffusion rate for poly(dimethylsiloxane). For polyacrylate-coated fibers the resistance might be higher, and thus diffusion rates lower and therefore more restrictive to the diffusion in the fiber itself. As a result of this, uptake rate constants in poly(dimethylsiloxane) could be limited by diffusion in the aqueous phase while in polyacrylate diffusion in the polymer material limits uptake (for log KSPME < 3.55). Data Analyses: QSPRs. In order to develop QSPRs, two approaches were followed: (i) PLS analyses using all type of quantum chemical descriptors in combination with log Kow as hydrophobicity parameter; (ii) MLR analysis using descriptors representing hydrophobic interaction as well as hydrogen bonding, based on the linear solvation energy relationship (LSER) as proposed by Kamlet et al.23 and modified by Cramer et al. to TLSER.14 The values for the calculated descriptors are given in Appendix I. (22) Arthur, C. L.; Killam, L. M.; Buchholz, K. D.; Pawliszyn, J. Anal. Chem. 1992, 64, 1960. (23) Kamlet, M. J.; Doherty, R. M.; Abboud, J.-L. M.; Abraham, M. H.; Taft, R. W. J. Pharm. Sci. 1986, 75, 338.

r2 Q2 LVb

log k1

importancea

PRC

importancea

PRC

0.6574 0.0822 -0.1211 0.0895 -0.1154 0.0736 0.1618 -0.0666 0.0327 0.0700 0.2009

0.6309 0.6326 0.1141 -0.0911 0.1702 -0.1244 0.0539 0.0614 -0.0030 0.0017 0.4332 2.011

0.6030 0.0557 -0.1571 0.1547 -0.1319 -0.0295 0.1175 -0.0530 0.0153 -0.0174 0.0464

-0.4364 0.5073 0.0675 -0.1033 0.2574 -0.1243 -0.0189 0.0390 -0.0021 0.0007 -0.0939 0.4063

0.939 0.831 4

0.933 0.829 4

a Five most important variables printed in italics. b Number of latent variables.

Table 3. Results of the Theoretical Linear Solvation Energy Relationships (TLSER) Using Multiple Linear Regression

intercept log Kow LUMO Q+ r2 Q2 a

log k1 RCa

log KSPME RCa

-0.4536 0.5561 -0.3324 0.8650 0.93 0.90

-0.3027 0.6877 -0.3867 2.920 0.94 0.91

Regression coefficients.

Figure 4. Partition coefficients of organic chemicals to polyacrylatecoated fibers as a function of log Kow from this study (9) and phenol data from Dean et al.24 (O).

The outcome of the PLS analyses is given both as the autoscaled (importance) and (original) pseudoregression coefficients (PRC) in Table 2. Good models were developed for log k1 and log KSPME. Since log k2 is fairly constant over the measured range, there is no necessity to model this. Log KSPME is described by log Kow, LUMO, η, RE, and Q+. Here log Kow and RE can be regarded as variables describing the hydrophobic character of the analytes LUMO, η, and Q+ as the Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

4461

Appendix I name 2-butoxyethanol 1-butanol aniline 3-pentanol 2-nitroaniline nitrobenzene 1-hexanol quinoline 2-nitrotoluene N,N-dimethylaniline 2-allylphenol chlorobenzene 2-phenylphenol 4-chloro-3-methylphenol p-xylene trichloroaniline 1,3,5-trichlorobenzene 4-n-pentylphenol 2,4,5-trichlorotoluene

HOMO (eV)

LUMO (eV)

EN (eV)

η (eV)

µ (D)

RE (au)

MSA (Å2)

MV (Å2)

Q- (au)

Q+ (au)

-10.650 -10.930 -8.5223 -10.805 -9.0683 -10.562 -10.956 -9.1814 -10.171 -9.3324 -9.0162 -9.5612 -8.7309 -9.0514 -9.0615 -8.6304 -9.9213 -8.9018 -9.4753

2.6755 3.4643 0.6392 3.4884 -0.7937 -1.0679 3.5490 -0.4666 -1.101 0.4336 0.3597 0.1545 -0.0489 0.0930 0.4871 -0.1974 -0.4022 0.4370 -0.4355

3.99 3.73 3.94 3.66 4.93 5.81 3.70 4.82 5.59 4.45 4.33 4.70 4.39 4.48 4.29 4.41 5.16 4.23 4.96

6.66 7.20 4.58 7.15 4.14 4.75 7.25 4.36 4.58 4.88 4.69 4.86 4.34 4.57 4.77 4.22 4.76 4.67 4.52

2.3155 1.7100 1.5420 1.5593 5.2942 5.2393 1.7000 1.8783 5.0370 0.9309 1.1352 1.3065 1.3343 1.0728 0.0768 3.0229 0.0007 1.3138 1.4672

8.418 7.987 9.904 6.601 13.68 10.10 5.386 13.14 12.09 11.57 12.83 8.995 16.31 12.21 10.82 14.75 11.69 16.13 13.89

178.94 168.95 125.97 144.54 149.93 138.33 125.67 154.50 156.78 170.31 175.52 127.38 197.27 156.06 153.22 169.51 161.23 229.46 180.08

128.06 119.14 94.59 102.69 116.57 105.02 86.31 122.65 121.53 129.13 135.85 96.32 160.92 122.58 115.08 134.48 125.14 175.63 140.62

-0.520 940 -0.554 400 -0.854 487 -0.539 372 -0.648 774 -0.493 948 -0.542 500 -0.653 760 -0.504 287 -0.619 966 -0.481 777 -0.126 192 -0.481 261 -0.489 415 -0.184 637 -0.684 868 -0.188 812 -0.495 086 -0.259 279

0.323 321 0.315 700 0.380 055 0.319 686 0.351 049 0.110 692 0.314 800 0.153 713 0.116 273 0.146 929 0.333 847 0.105 376 0.330 341 0.335 638 0.104 691 0.327 861 0.058 246 0.336 174 0.099 445

ability to form a hydrogen bond by donating a hydrogen atom to or accepting electrons from polyacrylate. Log k1 is mainly influenced by hydrophobicity (log Kow and RE) and LUMO, EN, and η. The last three variables are highly correlated. So, log k1 is mainly determined by the hydrophobicity of the compound and some LUMO-related properties, describing probably the ability to form hydrogen bonds. The TLSER models that contain only hydrophobic and hydrogen-bonding properties are shown to be even better than the models based on PLS analysis. The models are given in Table 3. Although the fitted models using the TLSER do not show a higher r2 than the PLS models, the predictive power, expressed by Q2, is better. For both log KSPME and log k1, log Kow, LUMO, and Q+ are the main descriptors. So besides increasing hydrophobicity, a decreasing LUMO, i.e., the ability of the chemicals to accept electrons from the polyacrylate, increases the affinity for the coating. In addition, log KSPME is related to Q+, indicating that the donation of a hydrogen atom to the polyacrylate coating on the fiber determines the affinity for this coating. The relationship between log KSPME and log Kow is presented in Figure 4. This plot also includes literature data for a few

phenols, taken from Dean et al.24 It is obvious that polar chemicals (substituted anilines, phenols, nitrobenzenes) have a higher KSPME value than nonpolar chemicals (substituted aromatics, alcohols, ethers) with the same hydrophobicity, probably because hydrogen bonding increases their interaction with the fiber coating. On the basis of these results, it can be concluded that different relationships will be valid for different classes of chemicals. Therefore it should be pointed out that KSPME is not a very accurate predictor of Kow as was suggested by Dean et al.24 ACKNOWLEDGMENT The financial support of the Dutch Ministry of Housing, Spatial Planning and Environment, Project 94230302, and the Basque Government, Department of Education, Universities and Investigation, is gratefully acknowledged. This work was partially carried out in the framework of the EC project Fate and Effect Modeling Using Structure-Activity Relationships (FAME) under Contract ENV4-CT96-0221. Received for review July 25, 1996. Accepted September 1, 1996.X AC9607471

(24) Dean, J. R.; Tomlinson, W. R.; Makovskaya, V.; Cumming, R.; Hetheridge, M.; Comber, M. Anal. Chem. 1996, 68, 130.

4462 Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

X

Abstract published in Advance ACS Abstracts, October 15, 1996.