Polymers as Reference Partitioning Phase - ACS Publications

Apr 26, 2016 - were cross-validated through comparison of the estimates obtained via different reference polymers. These estimates were averaged. Unce...
2 downloads 4 Views 869KB Size
Download PDF
Article pubs.acs.org/ac

Polymers as Reference Partitioning Phase: Polymer Calibration for an Analytically Operational Approach To Quantify Multimedia Phase Partitioning Dorothea Gilbert,*,†,‡,∥ Gesine Witt,¶ Foppe Smedes,§ and Philipp Mayer† †

Department of Environmental Engineering, Technical University of Denmark, DK-2800 Kongens Lyngby, Denmark Department of Environmental Science, Aarhus University, P.O. Box 358, DK-4000 Roskilde, Denmark ¶ Faculty of Life Science, Environmental Technology, Hamburg University of Applied Sciences, DE-21033 Hamburg, Germany § Masaryk University, RECETOX, Kamenice 753/5, 62500 Brno, Czech Republic ‡

S Supporting Information *

ABSTRACT: Polymers are increasingly applied for the enrichment of hydrophobic organic chemicals (HOCs) from various types of samples and media in many analytical partitioning-based measuring techniques. We propose using polymers as a reference partitioning phase and introduce polymer−polymer partitioning as the basis for a deeper insight into partitioning differences of HOCs between polymers, calibrating analytical methods, and consistency checking of existing and calculation of new partition coefficients. Polymer−polymer partition coefficients were determined for polychlorinated biphenyls (PCBs), polycyclic aromatic hydrocarbons (PAHs), and organochlorine pesticides (OCPs) by equilibrating 13 silicones, including polydimethylsiloxane (PDMS) and low-density polyethylene (LDPE) in methanol−water solutions. Methanol as cosolvent ensured that all polymers reached equilibrium while its effect on the polymers’ properties did not significantly affect silicone−silicone partition coefficients. However, we noticed minor cosolvent effects on determined polymer−polymer partition coefficients. Polymer−polymer partition coefficients near unity confirmed identical absorption capacities of several PDMS materials, whereas larger deviations from unity were indicated within the group of silicones and between silicones and LDPE. Uncertainty in polymer volume due to imprecise coating thickness or the presence of fillers was identified as the source of error for partition coefficients. New polymer-based (LDPE−lipid, PDMS−air) and multimedia partition coefficients (lipid−water, air−water) were calculated by applying the new concept of a polymer as reference partitioning phase and by using polymer−polymer partition coefficients as conversion factors. The present study encourages the use of polymer−polymer partition coefficients, recognizing that polymers can serve as a linking third phase for a quantitative understanding of equilibrium partitioning of HOCs between any two phases.

A

Later, partitioning-based methods have been further developed with the objective to determine thermodynamic measures such as the freely dissolved concentration,7,8 fugacity,9 or chemical activity10−12 to better understand partitioning processes. Such measurement end points strictly require (i) thermodynamic equilibrium and (ii) the negligible depletion criterion to be fulfilled, i.e., that the polymer does not deplete the sampled phase.13 Furthermore, partitioning-based methods have been developed to control freely dissolved concentrations in laboratory tests, e.g., for analytical passive dosing to determine binding and speciation of hydrophobic organic chemicals at controlled freely dissolved concentrations.14,15

wide variety of polymers is applied in analytical chemistry for the enrichment and selective extraction of organic chemicals. Traditionally, the methods aimed at a complete analyte transfer from the sample to the sorbent, such as in solidphase extraction (SPE) using polymeric sorbent particles packed into cartridges or incorporated into membranes (Empore extraction disks). In the past decades, enrichment techniques that employ polymers as a partitioning phase have been developed. In 1990, Arthur and Pawliszyn1 introduced solid-phase microextraction (SPME) as a technique that builds on partitioning as working principle. Thereupon, several other polymer-based partitioning methods such as stir-bar sorptive extraction,2 immobilized liquid extraction (ILE), and thin-film extraction evolved, reviewed by Baltussen et al.3 and Seethapathy and Górecki.4 Not least, partitioning of chemicals into a polymer forms the basis for many passive samplers5 and chemical sensors.6 © XXXX American Chemical Society

Received: January 29, 2016 Accepted: April 26, 2016

A

DOI: 10.1021/acs.analchem.6b00393 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry Polymers for Partitioning Enrichment. Amorphous polymers, especially elastomers, are especially suited as partitioning phase because they allow the accommodation and mobility of solutes within the polymer network. They function thermodynamically as a liquid phase in which a chemical species can diffuse. Elastomers have a very low glass transition temperature and consist of flexible chainlike molecules that facilitate the accommodation of solutes. Indeed, partitioning of organic chemicals into, for example, silicone rubber has been described as an absorption process.16,17 Elastomers possess a wide range of affinities for organic chemicals depending on their chemical structure.18 Their sorption capacity has been shown to depend on the spatial arrangement of the partitioning domains (amorphous parts) and the degree of cross-linking.19 Silicone polymers are among the polymers most frequently used in partitioning enrichment methods because of their suitable physical properties, thermostability, chemical inertness, and compatibility with various matrices.20 Furthermore, silicones were shown to be of a structure that facilitates the diffusion of hydrophobic organic chemicals within the polymer.21 Polydimethylsiloxane (PDMS) is one of the silicones most frequently applied in analytical chemistry.4 Another important polymer is polyethylene, especially low-density polyethylene, which is mainly used for sampling of chemicals from environmental matrices.22 Equilibrium Partitioning. Partitioning of a chemical species between phases is a process driven by differences in chemical potential between the phases until these differences are balanced, i.e., chemical equilibrium is reached. At infinite bath conditions, maximum enrichment of a chemical in a polymer can be achieved at equilibrium (equilibrium passive sampling). For two phases i and j in equilibrium, the measurement of a chemical’s state variable in one phase, e.g., the concentration in a polymer, can be used to deduce the chemical’s thermodynamic properties in the other partitioning phase. This is because thermodynamic laws dictate that at equilibrium the chemical potential μ of a compound is identical in all phases: μi = μj. Since the chemical potential is linked to chemical activity (a, unitless) in a logarithmic function, μ = μ° + ln a

K ij =

γjVmj Ci Z = = i Cj γiVmi Zj

(2)

where Vmj is the molar volume of phase j and Vmi is the molar volume of phase i. Hence, for phases in equilibrium, it is possible to project concentrations, activity coefficients, or fugacity capacities from one phase to another with knowledge of the partition coefficient Kij, an essential tool to understand chemical fate in multimedia systems. For a large number of organic chemicals, experimentally determined polymer-based partition coefficients can be found in the literature, mostly for environmental sampling phases like water and air. However, for a single analyte and one given polymer, partition coefficients can range over several orders of magnitude.26 This illustrates that determining accurate and precise partition coefficients for phases of very different capacity for a solute is challenging, especially for highly hydrophobic compounds. Di Filippo and Eganhouse26 critically assessed the quality of silicone−water partition coefficients from the SPME literature and pointed out that variation of these partition coefficients largely resulted from methodological errors but may also be due to uncertainty in fiber coating thickness and polymer source. Indeed, detailed polymer specifications are often not available from the suppliers. Rusina et al.27 compared polycyclic aromatic hydrocarbon (PAH) sorption of passive sampling polymers, and their study revealed that polymer−water partition coefficients for low-density polyethylene (LDPE), EXACT (ethylene/octane copolymer), and PDMS differed by up to 1.3 log units for selected polycyclic aromatic hydrocarbons (PAHs), whereas within a group of silicone rubbers the partition coefficients differed by only at most 0.4 log-units. These ranges show major sorption differences between polymer classes and indicate that also within a polymer class, e.g., within the class of silicone polymers, the sorption capacity of polymers can vary (here, we define a polymers’ sorption capacity as a property that is directly proportional to the fugacity capacity, i.e., the inverse activity coefficient). Polymers as a Reference Partitioning Phase. Ideally, one would always use the same polymer as sampling phase in all devices, which would allow the use of tabulated reference partition coefficients. Moreover, it would facilitate comparing and linking measured concentrations in passive samplers on a polymer basis. In consideration of the multitude of polymerbased measurement devices and passive samplers available, it is common practice to determine new material-specific partition coefficients through experiments, which however for environmental phases like water or air is not trivial. Such partition coefficients are difficult to determine experimentally because the partitioning phases involved exhibit extremely different solvation properties for hydrophobic organic chemicals (HOCs). In this case, a polymer can offer intermediate solubility. We suggest a complementary approach to deriving new polymer-specific partition coefficients (Kpol:phase) by using polymer−polymer partition coefficients as conversion factors to existing polymer partition coefficients Kref:phase (eq 3).

(1)

thermodynamic measures related to chemical activity such as the activity coefficient (γ = a/x with x being the mole fraction) or the fugacity ( f = a/(VmγZ) with Z being the fugacity capacity (mol/L Pa)) appear to be useful to describe and understand phase partitioning phenomena. Fugacity (f in units of Pa) is the partial pressure that a substance exerts into an ideal gas and is proportional to the concentration by the fugacity capacity: f = C/Z. Both chemical activities and fugacities are useful parameters to determine chemical equilibrium and form the basis for thermodynamic environmental chemical fate models.23−25 Polymer-Based Partition Coefficients. Commonly used partition coefficients Kij reflect the concentration ratio of a solute in two phases i and j that are in equilibrium. However, partition coefficients can also be understood thermodynamically as the inverse ratio of the molar activity coefficients or as the ratio of the fugacity capacities of two partitioning phases i and j:

K pol:phase = K ref :phase × K pol:ref

(3)

Polymer−polymer partition coefficients quantify differences in the sorption capacity of polymers and can be experimentally determined with very high accuracy and precision. Moreover, multimedia partition coefficients can be calculated on the basis B

DOI: 10.1021/acs.analchem.6b00393 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

partitioning phase, facilitating a multimedia phase partitioning understanding.

of individual polymer-based partition coefficients: the polymer as third partitioning phase then merely serves as a connective linking phase (Figure 1): K phase(x):phase(y) = K phase(x):ref /K phase(y):ref

■

MATERIAL AND METHODS Chemicals. PAHs (naphthalene, acenaphthene, fluorene, phenanthrene, anthracene, fluoranthene, pyrene, benz(a)anthracene, chrysene, and benzo(a)pyrene), polychlorinated biphenyls (PCBs; IUPAC congeners 3, 28, 31, 40, 44, 49, 52, 99, 101, 105, 110, 118, 128, 138, 149, 151, 153, 156, 170, 180, 187, 188, 194, 198, 209), and organochlorine pesticides (OCPs; α-HCH, β-HCH, γ-HCH, HCB, o’p-DDT, o’p-DDE, p’p-DDT, p’p-DDE, p’p-DDD, trans-nonachlor) were obtained in purity ≥98% (detailed information, including abbreviations, can be found in Table S2). The solvents methanol (≥99.9%), ethyl acetate (≥99.5% for analysis), and iso-octane (≥99.5% for analysis) were all purchased from Merck (Darmstadt, Germany); acetone from Rathburn Chemicals Ltd. Ultrapure water (Millipore, MA, USA) was used. Polymer Equilibration in Methanolic Solution. In total, 14 polymer materials were included in the study (Table S1), comprising silicone sheets, silicone coatings, silicone tubing, and silicone-coated SPME fibers as well as lay-flat LDPE tubing. All these materials have been reported suitable or potentially suitable as passive sampling materials. Further, silicone materials that proved applicable for passive dosing28,29 were included, i.e., medical-grade silicone elastomer specified to be pure PDMS and silicone O-rings, as well as silicone rods.15 The pure PDMS membrane from Specialty Silicone Products Inc. (in the following referred to as SSP), Altesil silicone sheets and/or silicone O-rings, and the silicone elastomer MDX44210 were chosen as reference materials for cross-validation of determined partition coefficients between experiments (Table S4). Polymers were prepared and solvent-cleaned as detailed in the Supporting Information. The exact weight or volume of each polymer was recorded. Polymers were equilibrated in a methanol−water solution. The geometry does not always allow the polymers to equilibrate in direct contact with each other. A methanolic solution was therefore chosen so that an equilibrium would establish between each polymer with the solution and thus an equilibrium between the different polymers. To introduce the chemicals, one spiked polymer material was added to each setup. A 60:40% v/v methanol/water ratio was chosen to facilitate a fast mass transfer of the chemicals and to avoid floating of the polymers. Silicone polymers and LDPE only have minimal absorption capacity for methanol:27 In a mixed solution of 60% methanol and 40% water, the fraction of methanol in the polymers was estimated to be not greater than 2% in silicone polymers (≤0.06% in LDPE). Given that methanol−silicone partition coefficients of HOCs are rather low,29,30 the mass fraction of excess analyte in the polymer due to methanol should be below 10% and will thus not have a large effect on polymer−polymer partition coefficients. A summary of the partitioning experiments is provided in Table S4. Briefly, polymers were placed in vials or jars coated with silicone and coexposed in 60/40% v/v methanol/water solution at 20 °C on a shaker. Equilibration times of 10 days for PAHs, 3 months for organochlorine chemicals in silicone polymers, and 6 months for organochlorine chemicals in LDPE were deemed long enough. On the basis of reported diffusivities by Rusina et al.21 and the maximal diffusion path length of 2 mm for silicone polymers and 0.07 mm for LDPE, internal diffusive equilibration was estimated to require only a few days.

(4)

Figure 1. Concept maps on the application of polymers as reference partitioning phase.

Such calculated partition coefficients may be superior not only for quantifying environmental multimedia partitioning but also for determining partitioning behavior of HOCs between miscible phases. Multiplication of polymer-based partition coefficients may then be less erroneous than the direct experimental determination of a partitioning coefficient. Aims. The general aim of the study is to establish and apply a new framework in which a polymer serves as a reference partitioning phase. More specific aims are to (1) systematically investigate partitioning of hydrophobic organic chemicals between various elastomers, i.e., silicone polymers and LDPE, that are applied in polymer-based analytical enrichment techniques, (2) apply polymer−polymer partition coefficients as conversion factors to derive new polymer-specific partition coefficients based on existing data, and (3) derive multimedia partition coefficients by using a polymer as a reference C

DOI: 10.1021/acs.analchem.6b00393 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

set to 1 min. The temperature program was as follows: 90 °C (1 min), 90 to 180 °C ramped 25 °C/min (hold time 2 min), 180 to 238 °C ramped 1.5 °C/min (hold time 2 min), 238 to 280 °C ramped 3 °C/min (hold time 10 min). Chromatograms were quantified on the basis of internal standard signals and nine-point external calibration curves using Agilent’s Chemstation software. QA/QC Measures. The GC method for the analysis of PCBs and the OCPs is accredited (ISO 17025) and subject to regular in-house quality control measures and interlaboratory comparisons. It is used for regular monitoring of chemicals in various environmental matrices. Limits of quantification using this method were determined to be in the range of 0.22−0.35 ng/mL, except for CB-3 (8.8 ng/mL). The measured concentrations in the samples were typically 100-fold above the detection limit. For analysis of PAHs by HPLC-FLD, direct injection of polymer extracts or methanol−water samples, hence minimal sample manipulation, has earlier proved to provide data of very high precision.15 External HPLC standards showed excellent linearity in the concentration range from 0.1 to 1000 ng/mL. PAH concentrations in the samples were in the range of approximately 20−200 ng/mL. Solvent blanks were included in both HPLC and GC analyses, in which the analytes could not be detected or else were below the limit of quantification. Data Treatment. For the PAH data set, concentrations in the measurement replicates of each vial (n = 3) were first averaged (CV < 3%) and the true replicates (concentrations in polymers from separate vials) were used to determine polymer−polymer partition coefficients (eq 5). Polymer− polymer partition coefficients were determined as the ratio of the measured analyte concentrations in the polymers (Cpol) of each replicate vial and then averaged:

Due to the relatively high solubility and consequently lower polymer−solvent partition coefficients of HOCs in methanolic solution, the transport of HOCs through methanolic solution is sufficiently faster than in pure water, thus providing equilibrium within the chosen exposure times. Furthermore, previous studies have demonstrated that the transfer of analytes from a methanolic solution into silicone can be accomplished within days28,31,32 or weeks30 at 20 °C. In a first experiment (see Table S4), vials with casted MDX44210 PDMS silicone were spiked with PAHs. The PAHs were dissolved in methanol at each 10 mg/L, and 1 mL of this solution was added into each of the triplicate vials to allow partitioning of the PAHs into the silicone. This loading step was repeated after 72 h. The vials were then repeatedly washed shortly with ultrapure water to remove any residual loading solution. Triplicates of each silicone material were placed into the spiked vials and allowed to equilibrate. In a second experiment (experiment B), PCBs and OCPs were spiked into casted silicone (MDX4-4210). The chemicals were dissolved in iso-octane and then combined to a mixture with each analyte present at ca. 2 mg/L. The solution was up-concentrated 20fold by reducing the solvent under a gentle stream of nitrogen. One hundred microliters of this stock solution (4 μg of each analyte) was added onto the casted silicone of seven replicate vials. The solvent was evaporated off under a gentle stream of nitrogen, which was confirmed gravimetrically. Then, silicone polymers and LDPE were added to these vials and equilibrated in methanol−water solution. Finally, jars and vials with thin silicone coatings (SilasticA and DC1-2577 silicone) were each equilibrated with selected reference polymers (experiments C and D, respectively). The vessels were placed horizontally on a roller to ensure good contact of the methanol−water solution with the silicone coating. At the end of each experiment, the methanol−water solution was removed and polymers were collected, shortly rinsed with water to remove adhering methanolic solution, and blotted dry on lint-free tissue. Chemical Analysis. PAH concentrations were measured in both polymer extracts and methanol−water solutions. PAHs were extracted from MDX4-4210 silicone with twice 5 mL of methanol, and extracts were combined. All other polymers were extracted with an excess of methanol (using ca. a 50-fold of the polymer volume). The extraction efficiency was confirmed with a second extraction with a smaller volume of methanol, which yielded concentrations generally less than 5% of the first extract or below the limit of quantification. PAHs were analyzed by HPLC with a fluorescence detector (Agilent 1100 series, G1321A FLD); details are provided in the Supporting Information. For measurement of PCB and OCP concentrations, polymers were extracted with iso-octane using a volume equal to at least a 20-fold of the polymer volume. PCB-55 was added as recovery standard. The extractions were repeated with the same volume as in the first extraction, and the extracts were finally combined. Recoveries confirmed extraction efficiencies between 96% and 121% (average 103%). Aliquots of the combined extracts were transferred into GC-vials, and PCB-155 was added as internal standard for gas-chromatographic quantification. Samples were analyzed by a GC-μECD (Agilent 7890A) equipped with a precolumn (SGE Analytical Science) and dual columns (J&W DB-5: 60 m × 250 μm × 0.25 μm; DB-1701: 60 m × 250 μm × 0.25 μm) for separation. Hydrogen was used as mobile phase. Samples were injected at 270 °C (1 μL on each column, splitless); equilibration time was

n

K pol . x:pol . y =

∑i = 1 (Cpol.xi /Cpol.y) i

n

(5)

For the PCB data set, an outlier test was performed on the analyte concentrations in replicate polymers (n = 4 to 7) prior to calculation of partition coefficients. Values outside the 95% probability interval for a normal distribution with the given mean concentration and its standard deviation were excluded (177 out of 3570). Concentrations in replicate polymers showed relative standard deviations typically between 2% and 5%. β-HCH, o’p-DDT, and p’p-DDT data were excluded because relative standard deviations of concentrations in polymers between replicates were >10% for most of the polymers in all experiments. Polymers that were equilibrated in SilasticA-coated jars (experiment D, see Table S4) showed analyte concentrations that varied by more than 5% (relative standard deviation) between replicates, and resulting polymer− polymer partition coefficients are therefore shown as tentative results. The mass balance for each chemical in the individual experimental setups was calculated as the ratio of total recovered analyte mass and the initially spiked analyte mass. Concentrations in the methanol−water phase were measured in the PAH-experiment and else were estimated using published polymer−solvent partition coefficients.30 These coefficients were not available for DDT isomers and their metabolites (DDX), HCH isomers, TNC and CB-3, CB-40, CB-188, CB194, CB-198, and CB-209, for which then only the mass fraction in the polymer phase was calculated. Recoveries were D

DOI: 10.1021/acs.analchem.6b00393 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 2. (A) Visual representation of polymer−polymer partition coefficients in a heatmap with light colors indicating similar sorption capacity of polymers (spaces filled gray). (B) Example data are plotted against log Kow which allowed identification of identical polymers, polymers with fillers, and polymers of different chemical structure. Filled symbols = PAHs, open symbols = PCBs, and remaining symbols = OCPs.

complete mass balance was calculated, recoveries ranged between 87% and 119% (average 99%, CV 5%). For the remaining analytes, the mass balance was based on concentrations measured in the polymers (for details, see the Supporting Information). Polymer−Polymer Partitioning. Polymer−polymer partition coefficients were compiled for 13 silicone materials and LDPE for 10 PAHs, 7 OCPs, and 25 PCB congeners. The final data set includes 1988 experimentally determined polymer− polymer partition coefficients and 1074 calculated polymer− polymer partition coefficients (not considering the inverse 1/ Kpol:pol), which are tabulated in the Supporting Information (note that partition coefficients Kpol.x:pol.y are only listed once in either of the tables for polymer x or polymer y). Polymer− polymer partition coefficients were determined with high accuracy, as data from independent experiments were in good agreement, i.e., generally deviating not more than 10% from each other (Figure S1). Precision of the experimentally determined partition coefficients was very high with relative standard errors (RSE) < 5% for the majority (96%) of the data and RSEs not larger than 11%. Also calculated polymer− polymer partition coefficients exhibited excellent precision: RSEs were not larger than 8%, which demonstrates that multiplication of polymer−polymer partition coefficients can result in precise estimates. Note that uncertainty was also reduced by cross-validating calculated values via two different reference polymers (Figure S1). All polymer−polymer partition coefficients ranged between 0 and 0.83 log units (a factor of 6.8) depending on the polymer combination (Figure 2A). The highest values were reached for the silicone polymer DC1-2577 (0.105 < |log KDC:pol| < 0.834) and LDPE (0.0007 < |log KLDPE:pol| < 0.724) each in combination with any of the other polymers. Without these two materials, polymer−polymer partition coefficients spanned within 0 and 0.25 log units (a factor of 1.7 for the silicone

compared across all experiments for each analyte, and in case recoveries fell significantly below the overall median (deviation >5× the median absolute deviation [MAD]), the mass balance was deemed incomplete and data were excluded (i.e., DDEisomers, p’p-DDD, TNC, α-HCH, CB-3, and CB-40 in polymers equilibrated in SilasticA-coated jars). Polymer−polymer partition coefficients for the chosen reference polymers (KAltesil:SSP, KO−ring:MDX4−4210, KAltesil:ring, KAltesil:MDX4−4210) from separate experimental setups (excluding SSP−Altesil partition coefficient from SilasticA-coated jars) showed agreement within a 10% margin of deviation and thus were averaged. Derivation and Cross-Validation of Partition Coefficients. For materials that were not equilibrated together, partition coefficients were calculated as the product of polymer−polymer partition coefficients for the reference polymers Altesil, SSP and/or MDX4-4210, and silicone Orings (eq 6, given for Altesil as example). K poli:polj = K poli:Altesil × KAltesil:polj

(6)

Because polymers were equilibrated with at least two reference polymers, calculated polymer−polymer partition coefficients were cross-validated through comparison of the estimates obtained via different reference polymers. These estimates were averaged. Uncertainties of calculated polymer−polymer partition coefficients were estimated by error propagation of the input data (Supporting Information). Multimedia coefficients were calculated as the product of polymer−media partition coefficients (Kpol:med) from the literature whereby polymer− polymer coefficients then merely served as correction factor for differences between polymers (eq 4).

■

RESULTS AND DISCUSSION Mass Balance. Recoveries of PAHs ranged from 87% to 112% (average 102%, CV 7%). For PCBs and HCB, for which a E

DOI: 10.1021/acs.analchem.6b00393 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

PDMS materials and DC1-2577. Indeed, the material specifications for DC1-2577 reveal that the polymer contains phenyl groups attached to the siloxane backbone. Altesil in combinations with all other PDMS or PDMS-like materials also showed a rather wide span of log Kpol:pol across all analytes. This suggests that functional side groups other than methyl-groups (as in PDMS) may be attached to the siloxane backbone of the Altesil silicone polymer. Polymer−polymer partition coefficients when plotted against the octanol−water partition coefficient (log Kow) revealed a detailed picture of how polymers differ in their capacity to absorb HOCs (Figure 2B−III). For example, increasing KLDPE:PDMS with increasing log Kow indicates that LDPE can accommodate more hydrophobic solutes better than PDMS. Interestingly, Altesil-PDMS and Polymicro-PDMS partition coefficients increase with solute hydrophobicity for PAHs, but this is not the case for PCBs (Figure 2A). New Partition Coefficients. When polymer−polymer partition coefficients are applied as conversion factors, new polymer-specific partition coefficients can now be derived from existing experimentally determined partition coefficients (eq 3). Accuracy of such derived coefficients will largely depend on the input data, whose quality should be cross-validated. New polymer-based partition coefficients are listed in Table 1 and in Tables S18−S21 and include lipid−LDPE and PDMS−air partition coefficients. Jahnke et al.34 determined lipid−polymer partition coefficients for thin silicone sheets (SSP), but for

materials only). Three groups of polymer−polymer partition coefficients can be distinguished (Figures 2B): (1) Kpol:pol close to unity, (2) Kpol:pol showing a constant offset from zero for all analytes, and (3) analyte-specific Kpol:pol. Kpol:pol close to unity were those partition coefficients for polymer combinations of MDX4-4210, SSP and Nagasep hollow fibers, Altec silicone rods, and biomedical silicone tubing. Note that both MDX4-4210 and SSP are specified a pure PDMS polymer by the manufacturers. Our results thus suggest that Altec silicone rods, both white and translucent, Nagasep hollow fibers, and the silicone tubing consist of PDMS. Notably, for a combination of Altesil and the Polymicro-fibers, partition coefficients were close to unity but not for combinations of either Altesil or Polymicro fibers with identified PDMS-like materials (Figure 2A), indicating that the Altesil silicone polymer and the Polymicro silicone fiber coating may be similar in chemical structure, yet different from pure PDMS. For silicone O-rings and Biscasil, each in combination with PDMS or PDMS-like polymers, partition coefficients showed an offset of 0.1 to 0.2 log units from zero that was constant across all analytes for given polymer combinations (Figure 2B− II). Both silicone O-rings and Biscasil are composite materials. Silicone O-rings were analyzed by scanning electron microscopy and were shown to contain diatomaceous earth as filler in the polymer matrix (Figure S4), likely to improve the physical stability of the polymer. The rings also contain iron oxide, which gives the rings their red color. SEM analysis of Biscasil revealed the presence of CaCO3 in the polymer matrix. Biscasil is a material consisting of silicone coated onto glass fiber fabric; hence, the exact polymer volume is unknown. This may have caused a further offset of polymer−polymer partition coefficients and certainly contributed to uncertainty of measured HOC concentrations in Biscasil. In fact, such uncertainty is intrinsic for any volume-based measurements on thin polymer coatings. Also, the coating thickness of silicone-coated fibers has been shown to vary,33 which is also reflected in our polymer−polymer partitioning data set: (1) An offset of log Kpol:pol from zero can be observed for coated fibers, Biscasil and SilasticA in combination with polymers of defined mass. (2) Associated standard deviations for those partition coefficients are higher than standard deviations of polymer− polymer partition coefficients for polymers of known mass. For example, relative standard deviations of polymer−polymer partition coefficients for combinations of the two fibers and Biscasil (n = 106) were for 56% of the data >5%. In summary, partition coefficients involving polymers whose volume is uncertain (e.g., due to the presence of a filler) may deviate by a constant factor from the true value. Such deviation can be quantified by equilibrating such polymers with identical polymeric material of defined volume (or mass). Determined polymer−polymer partition coefficients can then be used as universal conversion factors. Third, there was a group of polymer−polymer partition coefficients that showed a nonconstant offset from zero. Across analytes, log Kpol:pol spanned over a wider range for given polymer combinations (Figure 2B−III), reflecting analytespecific partitioning. Such behavior is typical when the involved partitioning phases interact differently with the solutes, i.e., when activity coefficients of the compounds differ between polymers. As expected, this was the case for combinations of silicone polymers and LDPE. The present data clearly show that this is also the case for HOC partitioning between pure

Table 1. Calculated Polymer-Specific and Multimedia Partition Coefficientsa Klip:LDPEb (kg/kg) HCB CB28 CB31 CB44 CB49 CB52 CB99 CB101 CB105 CB110 CB118 CB128 CB138 CB149 CB151 CB153 CB156 CB170 CB180 CB187

4.52 12.95 14.53 18.95 16.50 18.96 16.22 18.38 17.66 18.25 16.00 18.94 19.50 22.33 19.66 18.39 18.74 20.78 19.73 18.29

Klip:DCc log KDC:wd (kg/kg) (L/kg) 8.98 8.64 8.65 5.69 8.41 7.92 12.08 11.93 10.81 9.01 14.65 7.80 12.36 10.58 10.09 17.08 16.87 13.14 19.66 14.55

5.25 5.79 5.74 6.20 6.15 6.11 6.63 6.53 6.81 6.64 6.70 7.21 7.11 6.95 6.88 6.98 7.08 7.50 7.29 7.14

log KPDMS:aire,h (L/kg)

log Klip:wf (L/kg)

log Kawg,h (L/L)

6.37 7.11

6.22 6.74 6.70 6.97 7.08 7.01 7.71 7.62 7.85 7.59 7.88 8.11 8.20 7.98 7.88 8.22 8.31 8.62 8.58 8.31

−1.44 −1.86

7.68 7.56 8.11 8.47 8.40 8.70 8.69

8.56 8.74 9.07 9.08 9.03

−2.14 −2.00 −2.06 −2.31 −2.19 −2.12 −2.09

−2.04 −2.16 −2.11 −2.20 −2.24

For 20 °C if not indicated otherwise. bKlip:LDPE = Klip:SSP × KSSP:LDPE (Klip:SSP from Jahnke et al.34). cKlip:DC = Klip:SSP × KSSP:DC (Klip:SSP from Jahnke et al.34). dlog KDC:w = log KAltesil:w + log KDC:Altesil (KAltesil:w from Smedes et al.30). elog KPDMS:air = log KLDPE:air + log KPDMS:LDPE (KLDPE:air from Khairy and Lohmann36); PDMS = MDX4-4210. flog Klip:w = log Klip:SSP + log KAltesil:w + log KSSP:Altesil (Klip:SSP from Jahnke et al.34 and KAltesil:w from Smedes et al.30). glog Kaw = log KLDPE:w − log KLDPE:air (KLDPE:w from Smedes et al.30 and KLDPE:air from Khairy and Lohmann36). hfor 20−25 °C. a

F

DOI: 10.1021/acs.analchem.6b00393 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 3. Comparison of partition coefficients from this study with partition coefficients from the literature. Partition coefficients were calculated as the product of polymer-based partition coefficients at 20 °C (air 25 °C) and specific polymer−polymer partition coefficients from this study. Open symbols = PCBs; filled symbols = PAHs.

3C) with the air−water partitioning data published by Schenker et al.40 and the PAH data from Ma et al.,41 which both rely on literature-derived values that underwent a least-squares adjustment. Note that we did not adjust the LDPE−water partition coefficients that Smedes et al. determined at 20 to 25 °C as temperature-adjusted KLDPE:w differ by only at most 0.17 log units from the original data. Calculated lipid−water partition coefficients agreed well with the partition coefficients at 37° reported by Quinn et al.42 (Figure 3B) after a temperature correction using the van’t Hoff relationship and partitioning enthalpies that were calculated on the basis of poly parameter linear free energy relationships (ppLFERs; details in the Supporting Information). In contrast, the lipid−water partition coefficients given by Geisler et al.43 are consistently lower than those that we calculated. Geisler et al. determined lipid−water partition coefficients for storage lipids based on a 37 °C pp-LFER estimate and extrapolated these values then to 21 °C.43 The observed offset may be explained by the fact that the calibration data set they used for the lipid− water partitioning pp-LFERs is based on PDMS−water partition coefficients predicted by pp-LFERs39 which are consistently by ca. 0.5 log units below the SSP−water partition coefficients that we used. Finally, the number of conversion steps influences accuracy and precision of calculated partition coefficients: Here, lipid−water partition coefficients were calculated as the product of three terms, namely, a polymer− water partition coefficient, a polymer−lipid partition coefficient, and a polymer−polymer partition coefficient as conversion factor. Clearly, one needs to be aware of the uncertainty that each term introduces to the final estimated value, which sets a limit to the accuracy of partition coefficients derived from polymer-based partition coefficients. Cosolvent Effects on Partition Coefficients and Future Directions. Although the polymer−polymer partition coefficients obtained from our experiments show a remarkably high precision, it remains to be investigated to what extent methanol as a cosolvent during the polymer−polymer equilibration had an impact on the partitioning properties of the polymers. A close analysis of our data in comparison to polymer−polymer partition ratios that were obtained as a product of published polymer−water partition coefficients30 (Figure S2) reveals that methanol indeed can affect the partitioning properties of polymers. While this effect is relatively small within the group of silicones, the effect seems to be more pronounced for LDPE−silicone partition coefficients: Comparing cosolvent

LDPE and other polymers, such lipid partition coefficients have not yet been determined. Conversely, polymer−air partition coefficients for a larger set of organic chemicals have so far only been determined for LDPE and polyurethane foam (PUF) but are lacking for silicone, although silicone has been suggested as a promising sampling phase for diffusive passive air sampling of volatile and semivolatile organic chemicals.35 We also calculated multimedia partition coefficients (Table 1) based on polymer-based partition coefficients, whereby the polymer phase then merely served as reference partitioning phase. Using such an approach, it is especially important that any differences in the polymers’ sorption capacity for HOCs be numerically considered, i.e., quantified through a polymer− polymer partition coefficient. We computed lipid−water partition coefficients Klip:w (Table S20) and air−water partition coefficients Kaw (Table S21) as a product of lipid−polymer34 and polymer−water30 or polymer−air36 partition coefficients, respectively, and corrected with the appropriate polymer− polymer partition coefficients for sorption differences between polymers (eq 4). Comparison to Literature Data. Polymer−polymer partition coefficients were in reasonably good agreement with polymer−polymer partition coefficients that can be deduced from published literature values30 (Figure S2). The derived partition coefficients KLDPE:w, Klip:w, and Kaw when compared with partition coefficients from the literature generally showed good agreement (Figure 3). This demonstrates that polymer− polymer partition coefficients can be applied to translate from one to another polymer-specific partition coefficient with high accuracy, and it supports the conceptual framework outlined in Figure 1. Calculated LDPE−water partition coefficients deviated typically by 0.2 log units from LDPE−water partition coefficients given by Choi et al.,37 Smedes et al.,30 and Fernandez et al.38 (Figure 3A). Notably, Fernandez et al. had obtained LDPE from a different source whereas all other LDPE partition coefficients were determined for LDPE from Brentwood Plastics. Also, our calculated PDMS−air partition coefficients, which are based on experimentally determined KLDPE:air and KLDPE:PDMS partition coefficients, differ not more than 0.3 log units from experimentally determined values (CB28, CB-52, CB-101, and HCB) given by Sprunger et al.39 We also estimated air−water partition coefficients from published LDPE−water partition coefficients30 and LDPE−air partition coefficients.36 Our data is in good agreement (Figure G

DOI: 10.1021/acs.analchem.6b00393 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

determination of freely dissolved aqueous concentrations. In the case of polymer−lipid partition coefficients, this will allow a translation of measured concentrations in passive samplers to equilibrium partitioning concentrations in lipids, which can be used for assessing the bioaccumulation potential of HOCs and to estimate the baseline toxic potential based on membrane lipid concentrations. PDMS−air partition coefficients are presented within this study in the hope that they will advance the development and application of silicone-based passive air samplers. Finally, improved quality of polymer-based partition coefficients can also contribute to a better understanding of multimedia phase partitioning processes.

derived polymer−polymer partition coefficients (60/40% v/v methanol/water) to polymer−polymer partition coefficients equilibrated in pure water,30 we calculated an average difference of 0.10 log units (s = 0.09; n = 260) for 5 silicone materials. For LDPE−Altesil partition coefficients, differences between cosolvent and pure water partition coefficients were on average 0.18 log units (s = 0.09, n = 65). However, not all of the observed deviation should be attributed to the possible cosolvent effect in the present study. Ideally, polymer−polymer partition coefficients should be determined without any solvent that can enter a polymer at a concentration that can alter a polymer’s properties. We chose methanol as cosolvent in water because it is one of the few solvents causing minimal polymer swelling and a solvent that is miscible with water. Polymer coequilibration in water is not recommended due to the long equilibration times required and a high risk of inhomogeneous solute distribution in the polymers.46 We suggest that future experimental determinations of polymer−polymer partition coefficients be performed by reducing the methanol concentration over the course of the experiment.14

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.6b00393. Tables: polymer−polymer partition coefficients for combinations of the 14 polymers and polymer-specific polymer−lipid partition coefficients. Figures: an enlarged version of Figure 2A, graphs for cross-validation of determined polymer−polymer partition coefficients, and an electron microscopy image of the material structure of Altec silicone O-rings. Additional text: detailed information on Materials and Methods, error propagation, and temperature-correction of partition coefficients. (PDF)

■

CONCLUSIONS Application. Polymer−polymer partition coefficients will aid in the calculation of multimedia partitioning coefficients such as for lipid−water, sediment−water, or water−air partitioning, which within environmental chemistry can help one to understand chemical fate and thermodynamic phase relationships. Including a polymer as a third phase may also be practical when determining partitioning coefficients for HOCs between miscible phases, e.g., mixed solvents, or in emulsions. The main advantage would therefore be that HOCs are easy to measure in a solid-state polymer phase, which would not mix with the other phases. Finally, polymer−polymer partition coefficients when applied as polymer-specific conversion factors to literature values allow for a consistency check of published partition coefficients. Such kind of cross-validation strategy is similar to what has been proposed in the “three solubility approach” by Cole and Mackay.44 Differences between partition coefficients that exceed the magnitude of the specific polymer−polymer partition coefficient will reveal inconsistencies and systematic errors. Polymer−polymer partition coefficients may therefore serve as a tool to scrutinize existing partition coefficients and in this way improve the overall quality of partitioning-based analytical methods and data. Implications. Availability, accuracy, and precision of partition coefficients are essential components for progress in research within analytical and environmental chemistry. Deriving partition coefficients for new materials via polymer− polymer partition coefficients will be equally accurate and precise as the base polymer media partition coefficient and will therefore have great value especially within passive sampling of organic chemicals. Just as new passive sampling materials widen the possibilities for the field application of passive samplers, the availability of partition coefficients can extend the extrapolation possibilities. For example, recently developed coated jars for passive sampling of sediment11,45 enabled one to target measurements at sediment porewater concentrations, which however required the availability of partition coefficients specific for the silicone polymer used. Such polymer-specific partition coefficients can now be easily obtained by applying polymer−polymer partition coefficients as conversion factors to existing data. In the case of polymer−water partition coefficients, this will facilitate a more accurate and consistent

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: +47 461 67 416. Present Address ∥

D.G.: Norwegian Geotechnical Institute (NGI), P.O. Box 3930 Ullevål Stadion, N-0806 Oslo, Norway. Phone: +47 461 67 416. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank Annegrete Ljungqvist for help with GCμECD analysis and Jan Timmner, TNO, Utrecht, The Netherlands, for producing the SEM images. The authors also thank Frank Wania for helpful comments on the manuscript. This study was funded by Aarhus University through a PhD scholarship. D.G. acknowledges additional support from US-SERDP (14 ER03-035/ER-2431), and F.S., from the Czech Ministry of Education (LO1214 and LM2011028).

■

REFERENCES

(1) Arthur, C. L.; Pawliszyn, J. Anal. Chem. 1990, 62, 2145−2148. (2) Baltussen, E.; Sandra, P.; David, F.; Cramers, C. J. Microcolumn Sep. 1999, 11, 737−747. (3) Baltussen, E.; Cramers, C. A.; Sandra, P. J. F. Anal. Bioanal. Chem. 2002, 373, 3−22. (4) Seethapathy, S.; Górecki, T. Anal. Chim. Acta 2012, 750, 48−62. (5) Mills, G. A.; Gravell, A.; Vrana, B.; Harman, C.; Budzinski, H.; Mazzella, N.; Ocelka, T. Environ. Sci.: Processes Impacts 2014, 16, 369− 373. (6) McGill, R. A.; Abraham, M. H.; Grate, J. W. CHEMTECH 1994, 24, 27−37. (7) Vaes, W.; Urrestarazu Ramos, E.; Verhaar, H.; Seinen, W.; Hermens, J. Anal. Chem. 1996, 68, 4463−4467. H

DOI: 10.1021/acs.analchem.6b00393 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

(44) Cole, J. G.; Mackay, D. Environ. Toxicol. Chem. 2000, 19, 265− 270. (45) Jahnke, A.; Mayer, P.; McLachlan, M. S.; Wickström, H.; Gilbert, D.; MacLeod, M. Environ. Sci.: Processes Impacts 2014, 16, 464−472. (46) Valderrama, J. F. N.; Baek, K.; Molina, F. J.; Allan, I. J. Environ. Sci.: Processes Impacts 2016, 18, 87−94.

(8) Poerschmann, J.; Zhang, Z. Y.; Kopinke, F. D.; Pawliszyn, J. Anal. Chem. 1997, 69, 597−600. (9) Horstmann, M.; McLachlan, M. Environ. Sci. Technol. 1992, 26, 1643−1649. (10) Legind, C.; Karlson, U.; Burken, J.; Reichenberg, F.; Mayer, P. Anal. Chem. 2007, 79, 2869−2876. (11) Reichenberg, F.; Smedes, F.; Jonsson, J. A.; Mayer, P. Chem. Cent. J. 2008, DOI: 10.1186/1752-153X-2-8. (12) Mayer, P.; Toräng, L.; Glæsner, N.; Jönsson, J. Anal. Chem. 2009, 81, 1536−1542. (13) Mayer, P.; Tolls, J.; Hermens, J. L. M.; Mackay, D. Environ. Sci. Technol. 2003, 37, 184A−191A. (14) Birch, H.; Gouliarmou, V.; Lützhøft, H.-C. H.; Mikkelsen, P. S.; Mayer, P. Anal. Chem. 2010, 82, 1142−1146. (15) Gouliarmou, V.; Smith, K. E. C.; de Jonge, L. W.; Mayer, P. Anal. Chem. 2012, 84, 1601−1608. (16) Baltussen, E.; Sandra, P.; David, F.; Janssen, H. G.; Cramers, C. Anal. Chem. 1999, 71, 5213−5216. (17) Mayer, P.; Vaes, W. H. J.; Hermens, J. L. M. Anal. Chem. 2000, 72, 459−464. (18) Goss, K.-U. Anal. Chem. 2011, 83, 5304−5308. (19) Guo, X.; Wang, X.; Zhou, X.; Kong, X.; Tao, S.; Xing, B. Environ. Sci. Technol. 2012, 46, 7252−7259. (20) Jahnke, A.; Mayer, P. J. Chromatogr. A 2010, 1217, 4765−4770. (21) Rusina, T. P.; Smedes, F.; Klanova, J. J. Appl. Polym. Sci. 2010, 116, 1803−1810. (22) Adams, R. G.; Lohmann, R.; Fernandez, L. A.; MacFarlane, J. K.; Gschwend, P. M. Environ. Sci. Technol. 2007, 41, 1317−1323. (23) Mackay, D. Environ. Sci. Technol. 1979, 13, 1218−1223. (24) Diamond, M. L.; Mackay, D.; Welbourn, P. Chemosphere 1992, 25, 1907−1921. (25) Mackay, D. Multimedia environmental models: The fugacity approach, 2nd ed.; CRC Press: Boca Raton, FL, 2001; 261 p. (26) DiFilippo, E. L.; Eganhouse, R. P. Environ. Sci. Technol. 2010, 44, 6917−6925. (27) Rusina, T. P.; Smedes, F.; Klanova, J.; Booij, K.; Holoubek, I. Chemosphere 2007, 68, 1344−1351. (28) Smith, K. E. C.; Dom, N.; Blust, R.; Mayer, P. Aquat. Toxicol. 2010, 98, 15−24. (29) Smith, K. E. C.; Oostingh, G. J.; Mayer, P. Chem. Res. Toxicol. 2010, 23, 55−65. (30) Smedes, F.; Geertsma, R. W.; van der Zande, T.; Booij, K. Environ. Sci. Technol. 2009, 43, 7047−7054. (31) Booij, K.; Smedes, F.; van Weerlee, E. M. Chemosphere 2002, 46, 1157−1161. (32) Lee, H.; Shim, W. J.; Kwon, J.-H. Sci. Total Environ. 2014, 470− 471, 1545−1552. (33) Witt, G.; Lang, S.-C.; Ullmann, D.; Schaffrath, G.; Schulz-Bull, D.; Mayer, P. Environ. Sci. Technol. 2013, 47, 7830−7839. (34) Jahnke, A.; McLachlan, M. S.; Mayer, P. Chemosphere 2008, 73, 1575−1581. (35) De Coensel, N.; Desmet, K.; Górecki, T.; Sandra, P. J. Chromatogr. A 2007, 1150, 183−189. (36) Khairy, M. A.; Lohmann, R. Environ. Sci.: Processes Impacts 2014, 16, 414−421. (37) Choi, Y.; Cho, Y.-M.; Luthy, R. G. Environ. Sci. Technol. 2013, 47, 6943−6950. (38) Fernandez, L. A.; MacFarlane, J. K.; Tcaciuc, A. P.; Gschwend, P. M. Environ. Sci. Technol. 2009, 43, 1430−1436. (39) Sprunger, L.; Proctor, A.; Acree, W. E., Jr; Abraham, M. H. J. Chromatogr. A 2007, 1175, 162−173. (40) Schenker, U.; MacLeod, M.; Scheringer, M.; Hungerbühler, K. Environ. Sci. Technol. 2005, 39, 8434−8441. (41) Ma, Y.-G.; Lei, Y. D.; Xiao, H.; Wania, F.; Wang, W.-H. J. Chem. Eng. Data 2010, 55, 819−825. (42) Quinn, C. L.; van der Heijden, S. A.; Wania, F.; Jonker, M. T. O. Environ. Sci. Technol. 2014, 48, 5920−5928. (43) Geisler, A.; Endo, S.; Goss, K.-U. Environ. Sci. Technol. 2012, 46, 9519−9524. I

DOI: 10.1021/acs.analchem.6b00393 Anal. Chem. XXXX, XXX, XXX−XXX