Prediction of Apolar Compound Sorption to Aquatic NOM Accounting

Jun 18, 2019 - Prediction of Apolar Compound Sorption to Aquatic NOM Accounting for NOM Hydrophobicity Using Aqueous Two-Phase Systems ...
0 downloads 0 Views 1MB Size
Download PDF
Article Cite This: Environ. Sci. Technol. 2019, 53, 8127−8135

pubs.acs.org/est

Prediction of Apolar Compound Sorption to Aquatic Natural Organic Matter Accounting for Natural Organic Matter Hydrophobicity Using Aqueous Two-Phase Systems Kun Liu,† Heyun Fu,† Dongqiang Zhu,‡ and Xiaolei Qu*,† †

Downloaded via BUFFALO STATE on July 17, 2019 at 03:58:05 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Jiangsu 210023, China ‡ School of Urban and Environmental Sciences, Peking University, Beijing 100871, China S Supporting Information *

ABSTRACT: The equilibrium partitioning of organic compounds to natural organic matter (NOM) plays a key role in their environmental fate as well as bioavailability. In this study, a prediction model for organic compound sorption to NOM was theoretically derived based on two-phase systems. In this model, the hydrophobicity of NOM was scaled by their partition coefficients in an aqueous two-phase system (KATPS) and that of organics was scaled by their octanol−water partition coefficients (KOW). The model uses only KATPS and KOW as variables. Coefficients in the model were determined using a data set including the organic carbon−water partition coefficient (KOC) of four polycyclic aromatic hydrocarbons (PAHs) sorption to 10 NOM samples collected from surface waters. The resulting model was validated using additional NOM samples and reference NOM, which suggested good prediction power for PAH sorption to aquatic NOM. The model performance was compared with commonly used linear free energy relationship models, and its applicability was discussed. Sorption behavior unexpected by this model is attributed to additional sorption mechanisms other than partitioning. Overall, this approach allows prediction of KOC for apolar organic compound sorption to aquatic NOM simply using their respective partition coefficients in two-phase systems based on a specific model.

■

INTRODUCTION The equilibrium partitioning of organic compounds between natural organic matter (NOM) and water is the key environmental process that controls their fate, transport, bioavailability, and toxicity.1−3 Sorption of apolar organic compounds to NOM can often be treated as a partitioning process, in which the organic carbon−water partition coefficient, KOC, is one of the most important parameters for environmental modeling and risk assessment.1,3,4 Experimental determination of KOC is time- and labor-intensive due to the huge number of chemicals and enormous heterogeneity of NOM, if not technically impracticable. Thus, it is in great need to develop reliable prediction models for KOC. Current practice for predicting KOC relies on models utilizing linear free energy relationships (LFERs).5−9 Early single-parameter LFER (sp-LFER) can be used to estimate KOC of a given organic compound based on its octanol−water partition coefficient (KOW) or solubility.1,10,11 Nevertheless, spLFER cannot adequately reflect the diversity of the NOM phase and is only applicable to structurally closely related organics.5,8,9 Polyparameter LFER (pp-LFER) was later developed to cover a wider compound variability by considering diverse interactions in sorption.5,6,8 It is a multiple linear regression approach that uses several solute descriptors © 2019 American Chemical Society

as variables to capture intermolecular interactions affecting sorption. It is able to predict KOC of a wide variety of organic compounds in one equation using normally five solute descriptors.5,6,8 In pp-LFER, the characteristics of the NOM phase can be reflected by the coefficients for solute descriptors (i.e., system parameters). Nevertheless, system parameters can only be fitted using experimental partition coefficients for a large number of chemicals to given NOM.8 So far, they were determined for a handful of NOM, including Pahokee Peat, Aldrich humic acid (AHA), Suwannee River fulvic acid (SR FA), and a set of soil organic matter.6,9,12,13 The system parameters differ significantly depending on the properties of the NOM.14 Thus, both sp- and pp-LFER cannot adequately address the diversity of NOM. In order to make accurate KOC predictions, it is necessary to take the diversity of NOM into account in the prediction model as a variable. To this end, the major challenge is to properly scale the hydrophobicity of NOM which controls the partitioning process. We recently developed a scale system for NOM Received: Revised: Accepted: Published: 8127

January 24, 2019 June 17, 2019 June 18, 2019 June 18, 2019 DOI: 10.1021/acs.est.9b00529 Environ. Sci. Technol. 2019, 53, 8127−8135

Article

Environmental Science & Technology hydrophobicity using a poly(ethylene glycol) (PEG)/potassium citrate aqueous two-phase system (ATPS) (see Figure S1 for the steps involved in determining the partition coefficients of NOM in an aqueous two-phase system).15 In this ATPS, the PEG phase is more hydrophobic than the potassium citrate phase, and NOM components partition between these two phases mainly based on the hydrophobicity. The results suggested that the partition coefficients of NOM in PEG/ potassium citrate ATPS (KATPS) correlated well with the elemental, structural, and thermodynamic indexes commonly used to characterize NOM hydrophobicity.15 Thus, KATPS is a reliable quantitative measure of NOM hydrophobicity that can be potentially utilized to incorporate the diversity of NOM into the prediction model. We now derive a prediction model for KOC of organic sorption to NOM using KATPS and KOW as variables based on two-phase systems and a dual-component concept. Theoretical Approach (Two-Phase System Model). In PEG/potassium citrate ATPS, NOM is physically separated into two components: the NOM fraction in the PEG phase and that in the potassium citrate phase (Figure S1). NOM fractions dissolved in the PEG phase and the potassium citrate phase are defined as the hydrophobic (f Hydrophobic) and the hydrophilic components (f Hydrophilic, and f Hydrophobic + f Hydrophilic = 1), respectively. Sorption of organic compound to NOM can be conceptually considered as the sum of its sorption to these two NOM components: qoc = fHydrophobic qHydrophobic + fHydrophilic qHydrophilic

K OC = qoc /Ce = fHydrophobic qHydrophobic /Ce = KATPS/(KATPS + 1.22)KHydrophobic

where Ce is the aqueous-phase concentration of the organic compound at sorption equilibrium and KHydrophobic is the organic carbon−water partition coefficient of the organic compound to the hydrophobic NOM component. Assuming that K Hydrophobic is similar for all NOM (assumption 4), on the basis of the LFER: log KHydrophobic = a log K OW + b

log K OC = log[KATPS/(KATPS + 1.22)] + a log K OW + b (6)

Equation 6 suggests that one can predict the sorption of the organic compound to NOM simply based on the partition coefficients of the organic compound and NOM in two-phase systems if the aforementioned assumptions can be met. In the present study, we examined this hypothesis using sorption of four polycyclic aromatic hydrocarbons (PAHs) to 15 NOM samples as well as six reference NOM. PAHs are a group widely distributed pollutants in the environment due to incomplete combustion processes.16 The KOC was experimentally determined using sorption isotherms by batch sorption experiments. The KATPS was determined by the partitioning of NOM in a PEG/potassium citrate ATPS as the previous study suggests that it is a good quantitative measure of NOM hydrophobicity.15 The coefficients in the two-phase system model were determined by a training data set including 10 randomly picked aquatic NOM samples. The resulting model was validated using an additional five aquatic NOM samples and six aquatic/terrestrial reference NOM. The model was then compared with commonly used LFER models and its applicability discussed. Our objective was to develop a specific model to quantitatively predict KOC for organic compound sorption to NOM using their respective partition coefficients in two-phase systems (i.e., KATPS and KOW).

(1)

■

MATERIALS AND METHODS Materials. Phenanthrene (PHEN, 98%), pyrene (PYR, 98%), anthracene (ANTH, ≥99%), and poly(ethylene glycol) (PEG, BioUltra, Mw 10 000) were purchased from SigmaAldrich, U.S.A. Fluoranthene (FLUO, 98%) was purchased from J&K, China. Analytical-grade potassium citrate tribasic monohydrate, citric acid monohydrate, sodium hydroxide, and hydrochloric acid were obtained from Sinopharm Chemical Reagent Co., Ltd., China. Deionized water (18.2 MΩ·cm resistivity at 25 °C) was produced by an ELGA system (PURELAB Ultra, ELGA LabWater Global Operations, U.K.) and was used for all the experiments. Suwannee River NOM (SR NOM, 2R101N), Upper Mississippi River NOM (UM NOM, 1R110N), Suwannee River fulvic acid (SR FA, 2S101F), Pahokee Peat FA (PP FA, 2S103F), and Leonardite HA (LEO HA, 1S104H) were provided by the International Humic Substances Society (IHSS, U.S.A.). A homemade HA (LN HA) was extracted from the surface soil (0−20 cm) collected from Shenyang, Liaoning Province in northeast China using a standard method.17,18 The NOM stock solution was prepared by dissolving a predetermined amount of NOM powder in deionized water. After adjusting the pH to ∼6.8, the solution

(2)

Assuming that KATPS = CPEG/Cs (assumption 3), where CPEG is the NOM concentration in the PEG phase and Cs is the NOM concentration in the salt phase: fHydrophobic = C PEGVPEG/(CsVs + C PEGVPEG) = KATPS/(KATPS + Vs/VPEG)

(5)

Plugging eq 5 into eq 4 yields

where q oc is the organic carbon normalized sorbed concentration of organic compound to the NOM, qHydrophobic is the organic carbon normalized sorbed concentration of the organic compound to the hydrophobic NOM component, which can be conceptually seen as the fraction in the PEG phase ( f Hydrophobic), and qHydrophilic is the organic carbon normalized sorbed concentration of the organic compound to the hydrophilic NOM component, which can be conceptually seen as the fraction in the potassium citrate phase (f Hydrophilic). Assuming that hydrophobic partitioning is the dominate sorption mechanism (assumption 1), the sorption efficiency of the organic compound to the hydrophobic NOM component is considered to be much higher than that to the hydrophilic NOM component (i.e., qHydrophobic ≫ qHydrophilic). Since f Hydrophobic is also significantly higher than f Hydrophilic in most cas es, w e assum e that f H y d r o p h o b i c q H y d r o p h o b i c ≫ f HydrophilicqHydrophilic (assumption 2). Equation 1 can be simplified to qoc = fHydrophobic qHydrophobic

(4)

(3)

(Vs/VPEG = 1.22 in this study.) Thus 8128

DOI: 10.1021/acs.est.9b00529 Environ. Sci. Technol. 2019, 53, 8127−8135

Article

Environmental Science & Technology

Table 1. Summary of Partition Coefficients of NOM in Aqueous Two-Phase Systems (KATPS) and Organic Carbon−Water Partition Coefficients (KOC) for PHEN, ANTH, PYR, and FLUO Sorption to NOMa sample

KATPS

log KOC PHEN (log KOW = 4.57)

1 2 3 4 5 6 7 8 9 10

2.05 1.75 1.62 2.13 2.33 1.30 1.37 1.67 1.80 1.77

3.8359 3.6225 3.6054 3.6821 3.7771 3.5869 3.5792 3.6877 3.7004 3.7207

11 12 13 14 15 UM NOM SR NOM SR FA LN HA LEO HA PP FA

1.52 2.11 2.02 1.96 3.52 1.49 1.28 1.73 5.57 7.18 0.91

3.5733 3.7333 3.6552 3.6572 3.9136 3.6664 3.6615 3.7405 5.0284 5.2411 4.2026

log KOC ANTH (log KOW = 4.68) Training Set 3.8357 3.7493 3.7986 3.9448 3.9945 3.6289 3.6795 3.7575 3.7656 3.7479 Validation Set 3.7439 3.9125 3.8728 3.7074 4.0512 3.7794 3.7591 3.8140 5.0211 5.0978 4.7491

log KOC PYR (log KOW = 5.13)

log KOC FLUO (log KOW = 5.23)

3.9859 3.9865 3.9200 4.0230 4.1670 3.8190 3.8187 3.8230 3.8722 3.8512

4.0334 3.9903 3.9673 4.0933 4.2170 3.8492 3.8934 3.8996 3.9372 3.8575

3.9194 4.0855 4.0939 4.0956 4.2173 4.0375 3.8102 3.9785 5.1673 5.1792 4.3799

3.9272 4.1217 4.0907 4.0609 4.2514 3.9104 3.9280 3.9971 5.3012 5.3289 4.5561

a

Samples 1−15 are natural aquatic NOM samples collected from surface waters in Nanjing, China (see Table S1 for sample details).

was filtered through a 0.45 μm membrane (Supor-450, Pall, U.S.A.). The concentration of the NOM stock solution was determined by a total organic carbon (TOC) analyzer (vario TOC, Elementar, Germany). The stock solutions were stored at 4 °C in dark before use. The natural water samples were collected from rivers and lakes in Nanjing, China. The detailed sample information is summarized in Table S1. All water samples were filtered through 0.45 μm membranes (Supor-450, Pall, U.S.A.) after being collected and concentrated using a nanofiltration system (150 Da, BONA-GM-19, Bonabio, China). KATPS and KOC measurements were carried out using concentrated water samples. Partition Coefficients in ATPS (KATPS). The PEG/ potassium citrate ATPS is made of aqueous PEG and potassium citrate solutions. The top phase is aqueous 50% (w/w) PEG solution, and the bottom phase is aqueous 40% (w/w) potassium citrate solution. The pH of the potassium citrate solution was adjusted to 7 using 40% (w/w) citric acid solution prior to use. Partition experiments were performed in 15 mL graduated centrifuge tubes with 4 mL of PEG, 4 mL of potassium citrate, and 2 mL of NOM solution (∼15 mg C/L). The systems were thoroughly mixed by a vortex shaker and continuously mixed in a shaker (ZQLY-180GF, Zhichu, China) at 150 rpm and 20 °C for 50 min. After centrifugation at 4500 rpm (relative centrifugal force 2149g, Cence H1850R, Xiangyi Instrument, China) for 5 min, both top and bottom phases were sampled for absorbance measurements at 270 nm using a UV-2700 UV−vis spectrophotometer (Shimadzu, Japan) in a 3 cm quartz cuvette. A control sample was carried out using the same treatment with the addition of 2 mL of deionized water instead of NOM solution. The background absorbances of PEG and potassium citrate solutions in the control sample at 270 nm were subtracted from the

measurements. The KATPS was calculated by the ratio of the absorbance of the top phase to that of the bottom phase. See Figure S1 for the workflow and Table S2 for the absorbance of the NOM and control samples. Batch Sorption Experiments. Batch sorption experiments were conducted using a negligible-depletion solid-phase microextraction (nd-SPME) method described in previous studies.19,20 The PHEN (54 mg/L), ANTH (53 mg/L), PYR (115 mg/L), and FLUO (96 mg/L) stock solutions were prepared by dissolving predetermined amounts of chemicals in methanol. The 40 or 22 mL glass vials equipped with poly(tetrafluoroethylene)-lined screw caps were fully filled with the NOM solution, followed by the injection of predetermined amounts of PAH stock solution. The total volume of methanol carrier injected was kept below 0.1% (v/v) to minimize the cosolvent effect. The vials were tumbled in the dark at room temperature for 3 days to ensure sorption equilibrium. Then, a glass optical fiber coated with poly(dimethylsiloxane) (1 or 0.5 cm in length, 0.132 or 0.066 μL in coating volume, Polymicro Technologies, U.S.A.) was added into each vial. The vial containing the fiber was tumbled for 1 day. Then, the fiber was retrieved from the solution, wiped with a tissue, and extracted with 200 μL of methanol for 4 h. The resulting methanol solution was analyzed by highperformance liquid chromatography (HPLC, Agilent 1100) using a 4.6 × 150 mm Zorbax Eclipse XDB-C18 column (Agilent Technologies). PHEN, ANTH, PYR, and FLUO were quantified at 254, 254, 270, and 280 nm, respectively, using a mobile phase of 10% water/90% methanol (v/v). Calibration curves were built separately from controls receiving the same treatment as the sorption samples but without NOM. They included at least six concentrations which covered the whole concentration range tested with linear correlation coefficients (R2) of at least 0.99. The sorbed PAH was determined by the 8129

DOI: 10.1021/acs.est.9b00529 Environ. Sci. Technol. 2019, 53, 8127−8135

Article

Environmental Science & Technology

Figure 1. Correlations between organic carbon−water partition coefficients (KOC) for (a) PHEN, (b) ANTH, (c) PYR, and (d) FLUO sorption to NOM and partition coefficients of NOM in aqueous two-phase systems (KATPS).

KATPS/(KATPS + 1.22)]. The correlation between log KOC and log[KATPS/(KATPS + 1.22)] was examined for each given PAH as shown in Figure 1. For all PAHs, linear correlations can be found between log KOC and log[KATPS/(KATPS + 1.22)]:

difference between PAH added and that in aqueous solution at sorption equilibrium.

■

RESULTS AND DISCUSSION Predicting KOC of PAH Sorption to NOM Using TwoPhase Systems. The sorption isotherms of four PAHs (i.e., PHEN, ANTH, PYR, FLUO) to NOM samples are plotted as organic carbon normalized sorbed concentration (qOC) against aqueous-phase concentration (Ce) at sorption equilibrium (Figures S2 and S3). Table 1 summarizes the KOC determined by the slope of the linear regression between qOC and Ce (R2 > 0.81) and the KATPS of NOM samples. The structural and elemental properties of NOM can be estimated based on KATPS using eqs 7 and 8:15

PHEN: log K OC = 2.0367 log[KATPS/ [KATPS + 1.22)] + 4.1507 n = 10, R2 = 0.6468

ANTH: log K OC = 3.0727 log[KATPS/(KATPS + 1.22)] + 4.5006 n = 10, R2 = 0.8415

(10)

PYR:

aromatic carbon (%) = 51.994 log KATPS + 11.123, R2 = 0.80 (7)

O/C = − 0.445 log KATPS + 0.870, R2 = 0.85

(9)

log K OC = 2.8193 log[KATPS/(KATPS + 1.22)] + 4.5784 n = 10, R2 = 0.6744

(8)

The aromatic carbon content of tested NOM ranges from 9% to 58%. For aquatic NOM, the aromatic carbon content varies in a range of 17−40%. The O/C of tested NOM ranges from 0.49 to 0.84. For aquatic NOM, the O/C varies in a range of 0.63−0.83. Overall, the tested NOM is heterogeneous in elemental and structural properties. The log KOC of PAHs (i.e., PHEN, ANTH, PYR, FLUO) sorption to NOM generally increases with increasing log KOW (Figure S4), consistent with the partitioning theory.21 As discussed in the theoretical approach, the hydrophobicity of NOM can be quantified by the fraction of the hydrophobic component in ATPS [i.e.,

(11)

FLUO: log K OC = 2.8535 log[KATPS/(KATPS + 1.22)] + 4.6336 n = 10, R2 = 0.6746

(12)

For a given PAH, the constant term in eqs 9−12 (defined as d, which is equal to a log KOW + b as suggested by eq 6) is expected to be a function of KOW. Indeed, its value follows the order of dPHEN < dANTH < dPYR < dFLUO, consistent with the order of their KOW. However, there is a discrepancy between the fitted equations and eq 6 derived in the theoretical 8130

DOI: 10.1021/acs.est.9b00529 Environ. Sci. Technol. 2019, 53, 8127−8135

Article

Environmental Science & Technology

Figure 2. Comparison between experimental log KOC and model predicted log KOC for PAHs (i.e., PHEN, ANTH, PYR, and FLUO) sorption to 10 aquatic NOM samples in the training data set using (a) eq 13 and (b) eq 16, respectively. The dashed line represents the 1:1 line.

approach. Equation 6 suggests the coefficient for log[KATPS/ (KATPS + 1.22)] should be unity, while the fitted value is not. The coefficients in eq 6 can be fitted using the training data set:

log[(C LEO HA /CSR NOM)/((C LEO HA /CSR NOM) + 1.22)] = 1.40 log[(ALEO HA /A SR NOM )/((ALEO HA /ASR NOM ) + 1.22)] n = 15, R2 = 0.9792

(15)

In this case, the parameter c equals 1.40. The coefficients in eq 14 can be determined by the multiple linear regression:

log K OC = log[KATPS/(KATPS + 1.22)] + 0.4008(± 0.0463) log K OW + 2.1088( ±0.2275) n = 40, adjusted R2 = 0.663

log K OC = 2.7108( ±0.3161) log[KATPS/(KATPS + 1.22)]

(13)

+ 0.4008(± 0.0351) log K OW + 2.5041( ±0.1871)

The fitting result using eq 6 is not good with R = 0.663 and root mean squared errors (RMSE) of 0.08 (Figure 2a). We speculate that this stems from assumption 3 (i.e., KATPS = CPEG/Cs), which is not fully valid in our case. In this work, KATPS is experimentally determined by dividing A270 of NOM in the PEG phase by A270 of NOM in the salt phase (KATPS = APEG/As). This spectroscopic method does not consider the distribution of nonchromophores and was influenced by the fractionation of NOM in ATPS.15,22,23 Thus, the spectroscopically determined KATPS can be used to scale the NOM concentration ratio between these two phases but does not equal to CPEG/Cs.15 On the basis of the fitting results in Figure 1, we rewrite the eq 6 to 2

n = 40, adjusted R2 = 0.838

(16)

As shown in Figure 2b, the fitting quality is significantly improved with adjusted R2 = 0.838 and RMSE of 0.06. These results suggest that the sorption of PAHs to different aquatic NOM can be potentially predicted solely based on their partition coefficients in two-phase systems (i.e., KATPS and KOW). The fitted value of parameter c, 2.71, suggests that the optical differences between the hydrophobic and hydrophilic fractions of NOM in ATPS are even greater than those between LEO HA and SR NOM. Validation of the Two-Phase System Model Using Aquatic NOM Samples and Reference NOM. The twophase system model established above was validated using additional five NOM samples collected from surface waters as well as six reference NOM. The predicted KOC for PHEN, ANTH, PYR, and FLUO sorption to NOM in the validation data set was calculated using eq 16 and plotted against the KOC determined by batch sorption experiments (Figures 3 and 4; isotherms can be found in Figure S3). The plot for aquatic NOM samples and aquatic reference NOM (UM NOM, SR NOM, and SR FA) in the validation data set is shown in Figure 3. The aquatic reference NOM was collected from rivers in the United States and includes NOM fractions (i.e., SR FA). Consequently, the validation data set represents higher structural heterogeneity than NOM in the training data set. Consistently, the aromatic carbon contents of NOM in the validation data set varies from 19% to 40%, covering a broader range than that of the training data set, 17−30%. The O/C range of the validation data set is 0.63−0.83, which is also broader than that of training data set, 0.71−0.82. One may expect larger variation of KOC and more challenge to the model in this case. Nevertheless, the experimentally determined KOC for aquatic NOM is in good agreement with the predicted values (Figure 3). The prediction has an RMSE of 0.08, which is significantly lower than the reported values with pp-LFER

log K OC = c log[KATPS/(KATPS + 1.22)] + a log K OW + b (14)

in which c equals log[(CPEG/Cs)/((CPEG/Cs) + 1.22)]/ log[(APEG/As)/((APEG/As) + 1.22)]. In order to have a general understanding of the parameter c, let us assume that the hydrophobic fraction of NOM in the PEG phase has optical properties similar to LEO HA, a hydrophobic organic matter (KATPS = 7.18), and assume the hydrophilic fraction of NOM in the salt phase has optical properties similar to SR NOM, a hydrophilic organic matter (KATPS = 1.28). For each combination of LEO HA and SR NOM concentrations, CLEO HA/CSR NOM and ALEO HA/ASR NOM can be calculated separately. The absorbance data of LEO HA and SR NOM at different concentrations and calculated CLEO HA/CSR NOM and ALEO HA/ASR NOM are summarized in Table S3. ALEO HA/ ASR NOM is always higher than CLEO HA/CSR NOM, suggesting that hydrophobic NOM contains more chromophores than hydrophilic NOM. A good linear correlation was found between log[(C LEO HA /C SR NOM )/((C LEO HA /C SR NOM ) + 1.22)] and log[(ALEO HA/ASR NOM)/((ALEO HA/ASR NOM) + 1.22)]: 8131

DOI: 10.1021/acs.est.9b00529 Environ. Sci. Technol. 2019, 53, 8127−8135

Article

Environmental Science & Technology

(Figure 5). The sp-LFER model was calibrated by linearly fitting the correlation between log KOC and log KOW using the same training data set (Figure S4): log K OC = 0.4105( ±0.0607) log K OW + 1.8277( ±0.2983) n = 40, R2 = 0.5339

(17)

The pp-LFER model was adapted from the literature: log K OC = 0.54L − 0.98S − 0.42A − 3.34B + 1.20V + 0.02

6,9,28

ref 9

(18) log K OC = 0.81E − 0.16S − 0.20A − 2.34B + 2.11V + 0.25

ref 28

(19) log K OC = 0.88E − 0.53S + 0.88A − 2.16B + 2.54V − 0.05

ref 6

(20)

where L is the logarithm of the hexadecane/air partition constant of solute at 25 °C, E is the excess molar refraction, S is the polarizability/dipolarity parameter, A is solute H-bond acidity, B is solute H-bond basicity, and V is the McGowan molar volume. Solute descriptors for PAHs were adapted from the UFZ-LSER database (Table S4).29 As shown in Figure 5, the predicted log KOC of each PAH is the same for different NOMs based on LFER models. Thus, LFER models do not provide good prediction of log KOC, with RMSE ranges from 0.12 to 0.68, significantly higher than that of the two-phase system model, 0.08. A previous study also suggests that the best achievable RMSE of pp-LFER models is around 0.6−0.7.30 This is mainly due to the fact that LFERs do not adequately consider the diversity of NOM. This finding is consistent with the previous conclusion that pp-LFERs have unsatisfied performance in predicting the partitioning of organic compounds to complex materials such as humic substances.30 The results also indicate that the system parameters need to be carefully chosen if pp-LFER is used for sorption prediction. Considerations for Model Applications. The two-phase system model fits well with the conceptual model of equilibrium partitioning in environmental chemistry, i.e., it solves the equilibrium partition problem with laboratory equilibrium partition experiments. Its major advantage is that it addresses the diversity of the NOM phase as a simple variable in the equation for the first time. As a result, there is no need to calibrate the model for each NOM. Furthermore, it only requires two variables, KATPS and KOW, which can be determined using partition experiments or found in the literature. These experimentally determined values are generally considered to be more accurate than indexes estimated by models or software. The applicability of the two-phase system model needs to be further tested in terms of organic chemical properties, NOM properties, and solution chemistry. The two-phase system model is theoretically derived from mechanistic basis and a few sound assumptions. Thus, it is expected to be applicable for apolar organics whose sorption to NOM is controlled by partitioning. On the other hand, the applicability of the LFER models is defined by the diversity of the training data set. The impacts of solution chemistry on the sorption can be taken into account in the model using two approaches. Some of them can be addressed by changing the conditions of ATPS. For example, the KATPS of NOM at different pH values can be determined by adjusting the pH of the salt phase in ATPS. Thus, it is possible to involve pH effects into the model without adjusting coefficients. On the other hand, if the solution chemistry cannot be properly

Figure 3. Comparison between experimental log KOC and modelpredicted log KOC for PAHs (i.e., PHEN, ANTH, PYR, and FLUO) sorption to five aquatic NOM samples and three aquatic reference NOM (UM NOM, SR NOM, and SR FA) in the test data set. The dashed line represents the 1:1 line.

Figure 4. Comparison between experimental log KOC and modelpredicted log KOC for PAHs (i.e., PHEN, ANTH, PYR, and FLUO) sorption to three soil and peat reference NOM (LN HA, LEO HA, and PP FA) in the test data set. The dashed line represents the 1:1 line.

models, 0.39−1.30.9 This suggests that the two-phase system model has good prediction power for PAH sorption to aquatic NOM. Three terrestrial reference NOM from soil and peat origins were used to further test the applicability of the model. The model fails to predict the KOC for soil and peat reference NOM (LN HA, LEO HA, and PP FA) (Figure 4). A close examination of the data reveals that the KOC for soil and peat reference NOM are consistently higher than the predicted values. This indicates the presence of additional sorption mechanisms other than partitioning. Their strong sorption efficiency can be attributed to the presence of condensed organic matter domains or black carbon, which are commonly found in NOM extracted from peat and soil.24−27 These condensed structures will lead to sorption mechanisms other than partitioning, such as surface adsorption and micropore filling. Comparison with LFER Models for KOC Prediction. The prediction power of the two-phase system model was compared with that of the sp-LFER and pp-LFER models 8132

DOI: 10.1021/acs.est.9b00529 Environ. Sci. Technol. 2019, 53, 8127−8135

Article

Environmental Science & Technology

Figure 5. Comparison between experimental log KOC and predictions made by sp-LFER and pp-LFER models (eqs 17−20) for PAHs (i.e., PHEN, ANTH, PYR, and FLUO) sorption to five aquatic NOM samples and three aquatic reference NOM (UM NOM, SR NOM, and SR FA) in the test data set. The dashed line represents the 1:1 line.

decided by the properties of organic compounds, i.e., KOW. This is consistent with assumption 4. The two-phase system model has its limitations. Its applicability is mostly defined by the two assumptions made: (1) hydrophobic partitioning is the dominate sorption mechanism; (2) sorption of organic compound to the hydrophilic component of NOM is negligible. As a result, the model may not be applicable for polar organics that can have specific interactions with NOM such as hydrogen-bond and electrostatic interactions.2 On the other hand, the ppLFER model was known to work well for a wide range of

reflected by KATPS (such as ionic strength), the model will have to be retrained at the given conditions to get new model coefficients. The two-phase system model also provides information on the relative importance of the properties of the NOM and organic compounds. For all the NOM tested in this study, the NOM term (2.7108 log[KATPS/(KATPS + 1.22)]) is less important than the organic term (0.4008 log KOW). For highly hydrophobic NOM (i.e., KATPS ≫ 1.22), the NOM term approaches zero. In this case, the two-phase system model can be simplified to the sp-LFER model and the KOC is only 8133

DOI: 10.1021/acs.est.9b00529 Environ. Sci. Technol. 2019, 53, 8127−8135

Article

Environmental Science & Technology

(3) Luthy, R. G.; Aiken, G. R.; Brusseau, M. L.; Cunningham, S. D.; Gschwend, P. M.; Pignatello, J. J.; Reinhard, M.; Traina, S. J.; Weber, W. J.; Westall, J. C. Sequestration of hydrophobic organic contaminants by geosorbents. Environ. Sci. Technol. 1997, 31 (12), 3341−3347. (4) Weber, W. J.; McGinley, P. M.; Katz, L. E. A distributed reactivity model for sorption by soils and sediments. 1. conceptual basis and equilibrium assessments. Environ. Sci. Technol. 1992, 26 (10), 1955−1962. (5) Goss, K.-U.; Schwarzenbach, R. P. Linear free energy relationships used to evaluate equilibrium partitioning of organic compounds. Environ. Sci. Technol. 2001, 35 (1), 1−9. (6) Nguyen, T. H.; Goss, K.-U.; Ball, W. P. Polyparameter linear free energy relationships for estimating the equilibrium partition of organic compounds between water and the natural organic matter in soils and sediments. Environ. Sci. Technol. 2005, 39 (4), 913−924. (7) Zhu, D.; Pignatello, J. J. A concentration-dependent multi-term linear free energy relationship for sorption of organic compounds to soils based on the hexadecane dilute-solution reference state. Environ. Sci. Technol. 2005, 39 (22), 8817−8828. (8) Endo, S.; Goss, K.-U. Applications of polyparameter linear free energy relationships in environmental chemistry. Environ. Sci. Technol. 2014, 48 (21), 12477−12491. (9) Bronner, G.; Goss, K.-U. Predicting sorption of pesticides and other multifunctional organic chemicals to soil organic carbon. Environ. Sci. Technol. 2011, 45 (4), 1313−1319. (10) Karickhoff, S. W.; Brown, D. S.; Scott, T. A. Sorption of hydrophobic pollutants on natural sediments. Water Res. 1979, 13 (3), 241−248. (11) Chiou, C. T.; Porter, P. E.; Schmedding, D. W. Reply to comments by Maclntyre and Smith on: Partition equilibria of nonionic organic compounds between soil- organic matter and water. Environ. Sci. Technol. 1983, 17 (4), 227−231. (12) Neale, P. A.; Escher, B. I.; Goss, K.-U.; Endo, S. Evaluating dissolved organic carbon-water partitioning using polyparameter linear free energy relationships: Implications for the fate of disinfection by-products. Water Res. 2012, 46 (11), 3637−3645. (13) Endo, S.; Grathwohl, P.; Haderlein, S. B.; Schmidt, T. C. LFERs for soil organic carbon-water distribution coefficients (KOC) at environmentally relevant sorbate concentrations. Environ. Sci. Technol. 2009, 43 (9), 3094−3100. (14) Gschwend, P. M. Environmental Organic Chemistry; John Wiley & Sons: Hoboken, NJ, 2016. (15) Fu, H.; Liu, K.; Alvarez, P. J. J.; Yin, D.; Qu, X.; Zhu, D. Quantifying hydrophobicity of natural organic matter using partition coefficients in aqueous two-phase systems. Chemosphere 2019, 218, 922−929. (16) Samanta, S. K.; Singh, O. V.; Jain, R. K. Polycyclic aromatic hydrocarbons: environmental pollution and bioremediation. Trends Biotechnol. 2002, 20 (6), 243−248. (17) Sparks, D. L.; Helmke, P.; Page, A. Methods of Soil Analysis: Chemical Methods; Soil Science Society of America: Madison, WI, 1996. (18) Sun, H.; Shi, X.; Mao, J.; Zhu, D. Tetracycline sorption to coal and soil humic acids: An examination of humic structural heterogeneity. Environ. Toxicol. Chem. 2010, 29 (9), 1934−1942. (19) Fu, H.; Wei, C.; Qu, X.; Li, H.; Zhu, D. Strong binding of apolar hydrophobic organic contaminants by dissolved black carbon released from biochar: A mechanism of pseudomicelle partition and environmental implications. Environ. Pollut. 2018, 232, 402−410. (20) Bridle, H. L.; Heringa, M. B.; Schafer, A. I. Solid-phase microextraction to determine micropollutant-macromolecule partition coefficients. Nat. Protoc. 2016, 11 (8), 1328−1344. (21) Grathwohl, P. Influence of organic-matter from soils and sediments from various origins on the sorption of some chilorinated aliphatic-hydrocarbons - implications on KOC correlations. Environ. Sci. Technol. 1990, 24 (11), 1687−1693.

organic chemicals as it considers major solute−sorbent interactions. Our results suggest that the current model may not be directly applied to soil and peat NOM which induce additional sorption mechanisms. The model may not be applicable to highly hydrophilic NOM in which case assumptions 2 and 4 are not valid.

■

ENVIRONMENTAL IMPLICATIONS The two-phase system model allows us to evaluate partitioning variability that stems from both the compounds and NOM. It requires minimal information regarding the organic compound and NOM. The KOW of the organic compounds can be mostly found in the literature. One can estimate the KOC of apolar organic compounds sorption to a given aquatic NOM using the model by conducting a simple partitioning test of the NOM using ATPS. That will allow a more accurate and highresolution assessment of equilibrium partitioning and bioavailability of apolar organic compounds in aquatic systems. To this end, the model needs to be further validated using data sets covering a larger diversity of apolar organic compounds and NOM. The model also can serve as a tool to differentiate sorption mechanisms. The deviation from partitioning behavior predicted by the model can be used to probe additional sorption mechanisms and to reveal the characteristics of NOM. The two-phase system model is expected to find wide applications in understanding the fate and transport, bioavailability, and toxicity of apolar organics in aquatic systems, which warrant further investigation.

■

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.9b00529. Detailed information of the natural water samples, absorbance data in KATPS experiments, absorbance data of LEO HA and SR NOM at different concentrations, solute descriptors for PAHs, sorption isotherms of PAHs to NOM, and the correlation between log KOC and log KOW (sp-LFER model) using the training data set (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*Phone: +86-025-8968-0256; e-mail: [email protected]. ORCID

Heyun Fu: 0000-0002-0014-1829 Dongqiang Zhu: 0000-0001-6190-5522 Xiaolei Qu: 0000-0002-9157-4274 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the National Natural Science Foundation of China (Grants 21622703, 21876075, 21777002, and 21507056).

■

REFERENCES

(1) Chiou, C. T.; Peters, L. J.; Freed, V. H. A physical concept of soil-water equilibria for nonionic organic compounds. Science 1979, 206 (4420), 831−832. (2) Xing, B.; Senesi, N.; Huang, P. M. Biophysico-Chemical Processes of Anthropogenic Organic Compounds in Environmental Systems; John Wiley & Sons: New York, 2011; Vol. 4. 8134

DOI: 10.1021/acs.est.9b00529 Environ. Sci. Technol. 2019, 53, 8127−8135

Article

Environmental Science & Technology (22) Summers, R. S.; Cornel, P. K.; Roberts, P. V. Molecular-size distribution and spectroscopic characterization of humic substances. Sci. Total Environ. 1987, 62, 27−37. (23) Helms, J. R.; Mao, J. D.; Stubbins, A.; Schmidt-Rohr, K.; Spencer, R. G. M.; Hernes, P. J.; Mopper, K. Loss of optical and molecular indicators of terrigenous dissolved organic matter during long-term photobleaching. Aquat. Sci. 2014, 76 (3), 353−373. (24) Cornelissen, G.; Gustafsson, Ö .; Bucheli, T. D.; Jonker, M. T. O.; Koelmans, A. A.; van Noort, P. C. M. Extensive sorption of organic compounds to black carbon, coal, and kerogen in sediments and soils: Mechanisms and consequences for distribution, bioaccumulation, and biodegradation. Environ. Sci. Technol. 2005, 39 (18), 6881−6895. (25) Xing, B.; Pignatello, J. J. Dual-mode sorption of low-polarity compounds in Glassy Poly(Vinyl Chloride) and soil organic matter. Environ. Sci. Technol. 1997, 31 (3), 792−799. (26) Ran, Y.; Huang, W.; Rao, P. S. C.; Liu, D.; Sheng, G.; Fu, J. The role of condensed organic matter in the nonlinear sorption of hydrophobic organic contaminants by a peat and sediments. J. Environ. Qual. 2002, 31 (6), 1953−1962. (27) Pan, B.; Ghosh, S.; Xing, B. Nonideal binding between dissolved humic acids and polyaromatic hydrocarbons. Environ. Sci. Technol. 2007, 41 (18), 6472−6478. (28) Poole, S. K.; Poole, C. F. Chromatographic models for the sorption of neutral organic compounds by soil from water and air. J. Chromatogr. A 1999, 845 (1), 381−400. (29) Ulrich, N.; Endo, S.; Brown, T. N.; Watanabe, N.; Bronner, G.; Abraham, M. H.; Goss, K. U. UFZ-LSER Database, v 3.2 [Online]; Helmholtz Zentrum für Umweltforschung: Leipzig, Germany, 2017. (30) Stenzel, A.; Goss, K.-U.; Endo, S. Experimental determination of polyparameter linear free energy relationship (pp-LFER) substance descriptors for pesticides and other contaminants: New measurements and recommendations. Environ. Sci. Technol. 2013, 47 (24), 14204−14214.

8135

DOI: 10.1021/acs.est.9b00529 Environ. Sci. Technol. 2019, 53, 8127−8135