Development of Quantitative Structure–Activity Relationships for Nanoparticle Titanium Dioxide Aggregation in the Presence of Organic Contaminants

Chemical reagents

n-TiO₂ (Rutile) in powder form was purchased from Nanostructured and Amorphous Material, Inc. (Los Alamos, NM) with an average particle size of 10 × 40 nm, 98% purity, specific surface area of 160 m²/g, and coated with <5% SiO₂. Particularly, 12 representative OWCs were considered in this study due to the occurrence and persistence of those contaminants in various environments (Solomon et al., 1996; Konstantinou and Albanis, 2004; Ramesh and Maheswari, 2004; Lynch et al., 2006; Kemper, 2008; Swarcewicz and Gregorczyk, 2012; Petrie et al., 2015). The molecular structures and relevant decriptors of OWCs of interests, including pharmaceuticals (ibuprofen [Fluka], azithromycin [Fluka], and lincomycin [ICN Biomedicals, Inc.]), pesticides (alachlor [Fluka], butylate [Fluka], pendimethaline [Fluka], atrazine [Fluka], chlorthalonil [Riedel-de Haen], and cyanazine [Fluka]), and steroids (17β-estradiol [Aldrich], estrone [Aldrich], and 17α-ethinylestradiol [Sigma-Aldrich]), are provided in Table 1.

Preparation of n-TiO₂ suspensions with OWCs

n-TiO₂ suspensions were prepared as described in French et al. (2009), Chen et al. (2011), and Lee et al. (2015) with procedure modifications. All n-TiO₂ suspensions were prepared in Milli-Q water (18.0 MΩ-cm; Barnstead Nanopure System). About 0.3 mg of n-TiO₂ was placed in 300 mL of 2 mM CaCl₂ in a 500 mL glass flask resulting in a final n-TiO₂ concentration of 1 mg/L; these n-TiO₂ suspensions were immediately stirred using a magnetic stir-plate for 30 s followed by sonication in an ultrasonic water bath (FS 60, 100 W, 42 kHz; Fisher Scientific, Pittsburg, PA) for 1 h to obtain a homogeneous suspension. After sonication, the flask was cooled to 25°C and 0.01 M NaOH was added to adjust pH to 7 ± 0.1 and each OWC was separately spiked into the n-TiO₂ suspensions from a stock solution in methanol or Milli-Q water. A stock solution was prepared by <1% (v/v) to n-TiO₂ suspensions. The final concentration of methanol was kept <1% (v/v) to minimize the potential for cosolvent effects on n-TiO₂ aggregation. The final concentration of each OWC in the n-TiO₂ suspensions was 0.2 mg/L. Here, we fixed the concentration of n-TiO₂ and all OWCs with 1 and 0.2 mg/L, respectively even though those concentrations are higher than concentrations in actual environment because we considered the possible presence ratio between total suspended concentrations and representative OWCs concentration in effluent and surface waters that ranged from 0.0003 to 9.9 based on the information in our “Introduction” section.

Measurements of n-TiO₂ particle size and zeta potential

Aggregation measurements were conducted at 0 and 12 h to analyze aggregation sizes and zeta potential of n-TiO₂/OWC suspensions. For a representative sample collection, 3 mL aliquot was very carefully taken from n-TiO₂/OWC suspension by pipetting after vigorously shaking n-TiO₂/OWC suspension for seconds and n-TiO₂ aggregation particle size was measured by using a 90Plus particle size analyzer (Brookhaven Instruments Corporation, Holtsville, NY) based on dynamic light scattering (DLS). Measured particle size represents the effective diameter of NPs calculated for spherical particles from the Stokes-Einstein equation based on the translational diffusion coefficient, which is intensity weighted. Therefore, each effective diameter of NPs was an intensity-weighted hydrodynamic diameter from accumulated autocorrelation function. The mean diameter was calculated by the sum of these effective diameters divided by N runs (three runs in our experiments and 1 min per one run). Zeta potential of n-TiO₂ was measured by ZetaPALS (Brookhaven Instruments Corporation) using phase analysis light scattering technique to measure the electrophoretic mobility according to the Smoluchowski equation that converted mobility to the zeta potential of n-TiO₂. During the aggregation experiments, the flask containing the n-TiO₂ suspension with each OWC was covered with aluminum foil to prevent photo-degradation or photocatalytic degradation of the OWC by light and incubated in a dark chamber at room temperature.

To determine the rates of change in hydrodynamic diameter, the fast aggregated n-TiO₂ was measured before adding OWCs, shortly after (9 min) adding OWCs, and the slow aggregated n-TiO₂ was measured at an extended period (12 h) after adding OWCs. An aggregation rate kinetic (k _t ) was estimated by applying an aggregation kinetic equation published in previous studies (Mylon et al., 2004; Chen and Elimelech, 2006); \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} { k_t } = \left( { { \frac { \Delta { D_ { \rm { H } } } \left( t \right) } { \Delta t \cdot { N_0 } } } } \right) \tag { 1 } \end{align*} \end{document}

where k_t is the aggregation rate kinetic at time t, ΔD_H(t) is the difference between n-TiO₂ hydrodynamic diameter at time t and before adding OWCs, N₀ is the concentration of n-TiO₂, and, Δt is the time duration. Two time points, t = 9 min and 12 h, were only evaluated to represent the diameter change at the beginning and after an extended time duration that represent the aggregation rate kinetic of k₁ and k₂, respectively, which would demonstrate the fast aggregation and slow aggregation of n-TiO₂ in the presence of OWCs. For the convenience of discussion, the early time point (9 min) is denoted as 0 h in the “Discussion” section.

QSARs for aggregation of n-TiO₂ by OWC molecular descriptors

QSARs for predicting the aggregation of n-TiO₂ based on the OWC molecular descriptors were developed by using two different MLR models in MATLAB (R2015a). In model 1 (using fitlm function in MATLAB), it was developed with the following seven descriptors of OWCs: pK_a, solubility (C_w), log K_ow, molecular weight (M.W.), polar surface area (P.S.A.), molar volume (M.V.), and # of H Bond Donor (# of H.B.D.). In model 2 (using stepwiselm function in MATLAB), it was developed by forward and backward stepwise regression approaches by removing or adding OWCs descriptors in the model until its descriptors were satisfied with p-value <0.05.

Zeta potentials of n-TiO₂ in presence of OWCs

We measured the zeta potential of n-TiO₂ in the presence of OWCs to determine how n-TiO₂ surface charge is influenced by the adsorption of OWCs. The statistical significance of OWCs on zeta potential of n-TiO₂ at both 0 and 12 hr was evaluated using one-tailed Student's t-test assuming unequal variance (α = 0.05). Some OWCs such as azithromycin, ibupropen, chlorothalonil, alachlor, pendimethalin, and butylate resulted in statistically significant changes in the average zeta potential of n-TiO₂ when compared to control with no OWCs at 0 h. At 12 h, some other OWCs estrone, 17β-estradiol, chlorothalonil, atrazine, pendimethalin, and butylate showed statistically significant changes in the average zeta potential of n-TiO₂ when compared to control with no OWCs (Fig. 1). Particularly, OWCs satisfying with Student's t-test results have common functional groups of alkanes (R–CH₃) or hydroxyl (R–OH) except of chlorothalonil that are likely to provide a unique molecular to interact with hydroxylated n-TiO₂ aggregate surface (Table 1).

OPEN IN VIEWER

FIG. 1.

Zeta potential changes of n-TiO₂ in presence of OWCs (0.2 mg/L) and in absence of OWCs (control) in 2 mM CaCl₂ solution with pH 7 at 0 and 12 h. The mark of “*” presents the unsatisfied t-test results for the zeta potential changes. OWC, organic wastewater contaminant; n-TiO₂, nanoparticle titanium dioxide.

Adsorption tendency of OWCs is typically determined by sorption property of OWCs based on log K_ow that can be considered to the tendency of the chemical to partition with particulate phase. Chemicals with <1.0 may be considered relatively hydrophilic, while chemicals with >4 may be considered very hydrophobic. The log K_ow values of above OWCs ranged from 3.05 (chlorothalonil) to 5.18 (pendimethalin) and ranged from 2.61 (atrazine) to 5.18 (pendimethalin) in the t-test results at 0 and 12 h, respectively. These results demonstrated that those OWCs are able to easily partition on n-TiO₂ aggregate surfaces due to higher hydrophobicity, which may influence surface charge of n-TiO₂. Particularly, pendimethalin and butylate having higher log K_ow values of 5.18 and 4.15, respectively than other OWCs are very hydrophobic. As a result, these two OWCs could be more easily sorbed onto the n-TiO₂ aggregate surfaces and presumably the most negative surface charge changes are likely dominated by strong steric repulsion from pendimethalin and butylate adsorption.

The average zeta potential of n-TiO₂ was −5.2 mV (σ = ±1.0) at 0 h in the absence of OWCs, while the average zeta potential of n-TiO₂ showed more negative in the presence of some OWCs such as azithromycin (−7.7 mV), ibupropen (−7.1 mV), and alachlor (−8.4 mV). Particularly, the significant changes in zeta potential at 0 h were observed in the presence of chlorothalonil (−11.3 mV), butylate (−15.2 mV), and pendimethalin (−15.2 mV) based on t-test results (Fig. 1). In addition, the zeta potential of n-TiO₂ showed less negative in the presence of estrone (−4.9 mV) and 17β-estradiol (−4.5 mV), while more negative zeta potentials were observed in the presence of chlorothalonil (−9.1 mV) and atrazine (−9.4 mV) at 12 h. Moreover, the most negative zeta potentials were observed in the presence of pendimethalin (−16.7 mV) and butylate (−17.1 mV) (Fig. 1). These results showed clearly that the surface charge changes of n-TiO₂ were influenced by the adsorption of OWCs, which is consistent with prior studies (Domingos et al., 2009; Thio et al., 2011; Zhu et al., 2014). A few studies have reported increased stability of n-TiO₂ due to steric repulsion by the adsorption of organic contaminants (Pettibone et al., 2008; Joo et al., 2009; Mudunkotuwa and Grassian, 2010). A recent study demonstrated that the surface modification of n-TiO₂ by adsorption of critic acid resulted in more negative zeta potential changes at pH 6, which showed the effects of citric acid molecular properties on increasing stability of n-TiO₂ due to increased repulsive force (Mudunkotuwa and Grassian, 2010). In addition, a study showed the possibility of functional groups such as nonpolar carbon rings in 17β-estradiol on binding with the n-TiO₂ aggregate surfaces due to the presence of uncharged surface heterogeneity in n-TiO₂ (Lee et al., 2015).

Our results of zeta potential changes between at 0 and at 12 h in the presence of OWCs illustrated continuous interaction occurring between reactive n-TiO₂ surface heterogeneity and OWCs in time series. Moreover, the variation in our observed zeta potentials of n-TiO₂ with OWCs supported that unique properties of each OWC may play a key role on the surface modification of n-TiO₂ in the aspect of molecular interaction between n-TiO₂ surfaces and OWCs descriptors.

Aggregation behaviors of n-TiO₂ in presence of OWCs

The aggregation behaviors of n-TiO₂ in the presence of OWCs were presented by the average hydrodynamic diameters obtained from DLS measurements between 0 (fast aggregation) and 12 h (slow aggregation). The statistical significances of OWCs on average hydrodynamic diameters of n-TiO₂ at both 0 and 12 h were evaluated by one-tailed Student's t-test assuming unequal variance (α = 0.05). We observed that most of OWCs resulted in statistically significant changes in average hydrodynamic diameters of n-TiO₂ except for butylate (at 12 h) comparing to a control in the absence of OWCs, which showed the clear impact of OWCs adsorption on different n-TiO₂ aggregates.

The differences in hydrodynamic diameters of n-TiO₂ (Δ diameter) between in the presence and in the absence of each OWC at 0 and 12 h were shown in Fig. 2 and it presented that our target OWCs significantly affected the average hydrodynamic diameter changes of n-TiO₂. The average hydrodynamic diameter changes of n-TiO₂ in the absence of OWC were 381 nm for 0 h and 736 nm for 12 h that demonstrated the n-TiO₂ self-aggregation during times that may show the maximum hydrodynamic diameter changes of n-TiO₂ because our previous study showed the average hydrodynamic diameter of n-TiO₂ with 757 nm for 12 h in the same water chemistry conditions (only the concentration of n-TiO₂ is different with having 5 mg/L) (Lee et al., 2015). The largest Δ diameter (359 nm) was observed by chlorothalonil at the fast aggregation of n-TiO₂ (at 0 h), while the largest Δ diameter was observed by alachlor with 436 nm at the slow aggregation of n-TiO₂ (at 12 h). Particularly, we observed that Δ diameter changes (173 nm) of n-TiO₂ with butylate were larger than Δ diameter changes (135 nm) of n-TiO₂ in the presence of 17α-ethinylestradiol even though the zeta potential (−15.2 mV) in butylate was higher than the zeta potential (−4.9 mV) in 17α-ethinylestradiol at 0 h, which was the opposite results according to DLVO theory (Fig. 2). Δ diameter changes of n-TiO₂ decreased from 174 at 0 h nm to 53 nm at 12 h in the presence of butylate, however, Δ diameter changes of n-TiO₂ in the presence of pendimethalin increased from 153 nm at 0 h and 264 nm at 12 h even though the zeta potentials were similar between two OWCs. These results were in contrast with previous studies that increased zeta potential by adsorption of organic contaminants caused increases on repulsive electrostatic and steric repulsions, which resulted in increased stability of n-TiO₂ (Domingos et al., 2009; Mudunkotuwa and Grassian, 2010; Chen et al., 2012; Lee et al., 2015). However, our results may implicate that molecular properties of OWCs have strongly influenced the interaction with n-TiO₂ surfaces that resulted in different surface modification. Few studies reported the adsorption effects of organic compounds on the aggregation of n-TiO₂, however, some limited studies demonstrated different aggregation of n-TiO₂ by adsorption of oxalic acid and critic acid (Pettibone et al., 2008; Mudunkotuwa and Grassian, 2010). Particularly, Pettibone et al. (2008) demonstrated that the adsorption of oxalic acid on the surface of n-TiO₂ occurred because of the carboxylate group in oxalic acid and they implicated the role of specific physicochemical properties of contaminants on the aggregation of n-TiO₂. Therefore, our aggregation results could be explained by the effects of different affinities, unique physicochemical properties, and different functional groups of OWCs.

OPEN IN VIEWER

FIG. 2.

Diameter changes (Δ diameter) of n-TiO₂ in presence of OWCs (0.2 mg/L) and in absence of OWCs (control) in 2 mM CaCl₂ solution with pH 7 at 0 and 12 h. The mark of “*” presents the unsatisfied t-test result for the diameter changes.

From our aggregation experimental results, we determined n-TiO₂ aggregation rate kinetics between 0 and 12 h by using Equation (1). The aggregation rate kinetics of n-TiO₂ (k₁ and k₂ for 0 and 12 h, respectively) were shown in Fig. 3. At 0 h, the largest k₁ was observed in chlorothalonil (39.852 nm · L/[min · mg]), while the smallest k₁ was observed in cyanazine (9.374 nm · L/[min · mg]). After 12 h, the biggest k₂ was observed in alachlor (0.606 nm · L/[min · mg]), while the smallest k₂ was observed in butylate (0.073 nm · L/[min · mg]), which is the same results of Δ diameter in Fig. 2. Particularly, the aggregation rate kinetics between k₁ and k₂ among OWCs showed the r² values of 0.89 and 0.88, respectively. These observations suggested that the aggregation behavior of n-TiO₂ is strongly influenced by the adsorption of each OWC having unique physicochemical properties. In addition, the aggregation rate kinetics of n-TiO₂ between 0 and 12 h showed differences, demonstrating that interaction time influenced on the n-TiO₂ aggregation with OWCs in environmental systems.

OPEN IN VIEWER

FIG. 3.

Aggregation rate kinetics: k₁ for 0 h (A) and k₂ for 12 h (B).

QSARs for n-TiO₂ aggregation rate kinetics by OWCs descriptors

MLR was used to develop QSARs for predicting n-TiO₂ aggregation rate kinetics with seven descriptors of OWCs by MATLAB. Particularly, we developed QSARs for k₂ that may present the actual aggregation behaviors by slow interaction of n-TiO₂ with each OWC for 12 hr and the model 1 developed for k₂ was as follows: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} \begin{split} {\rm k}_2 & = 0.494 - 9.54 \cdot 10 ^{- 3} \cdot ( {\rm pK}_{\rm a} ) \\ & \quad + 1.59 \cdot 10 ^{- 4} \cdot ( {\rm C}_{\rm w} ,\hbox{water solubility} ) \\ & \quad - 0.0404 \cdot ( {\rm logK}_{\rm ow} ) + 8.01 \cdot 10 ^{- 4} ( {\rm M.W.} ) \\ & \quad - 1.88 \cdot 10^{- 4} \cdot ( {\rm P.S.A.} ) - 8.76 \cdot 10 ^{- 7} \cdot ( {\rm M.V.} ) \\ & \quad - 0.104 \cdot ( {\rm \ of \ H.B.D.} ) \end{split} \tag{2} \end{align*} %end \end{document}

For model 1, we observed positive coefficients for C_w and M.W., indicating that these properties can increase aggregation kinetics. Particularly, the two descriptors of C_w and M.W. may influence each other in the QSAR because the solubility of a compound tends to decrease with increasing M.W. according to Raoult's law in ideal conditions (Chiou, 2003). Other descriptors such as logK_ow, P.S.A. and # of H.B.D. are likely to relate with the polarity of OWCs inducing chemical reactivity (Su et al., 2012). Particularly, larger log K_ow and # of H.B.D. would result in less polar OWCs and could react with nonpolar surface areas n-TiO₂. In this QSAR, pK_a and M.V. had little effect on n-TiO₂ aggregation.

In addition, the model 2 developed for k₂ at 12 h was as follows: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} \begin{split}{\rm k}_2 = 0.598 - 0.0096 \cdot ( {\rm pK}_{\rm a} ) - 0.0688 \cdot (\hbox{log K}_{\rm ow} ) \\ \quad+ 0.0008 \cdot ( {\rm M.W.} ) - 0.1052 \cdot ( \hbox{of H.B.D.} )\end{split} \tag{3} \end{align*} \end{document}

For model 2, we observed that n-TiO₂ aggregation rate kinetics were related with four descriptors: pK_a, logK_ow, M.W., and # of H.B.D.. In both QSAR models, increased n-TiO₂ aggregation rates were likely related with M.W. of OWCs due to positive values, while other descriptors had negative values that may affect decreased n-TiO₂ aggregation. The log K_ow values of OWCs were related with the solubility, thus as the value of log K_ow decreases, it presents that the solubility increases in aqueous phase. The other descriptors, pK_a and # of H.B.D., present the polarity of OWCs that can induce chemical reactivity between n-TiO₂/OWCs. In addition, an organic compound having less acidic or small hydrogen bond donors would be less polar (Henriksen et al., 2005; Borges et al., 2017), therefore the polarity of OWCs depending on pK_a and/or # of H Bond Donors may play a key role on the aggregation of n-TiO₂. From the developed QSARs (model [2] and [3]), scatter plots between the developed QSARs versus the experimental aggregation rate kinetics were shown with a reference line in Figs. 4 and 5. The majority of the OWCs were located near the reference line with the exception of butylate that is likely related with the t-test results of diameter changes of n-TiO₂ in the presence of OWCs, which may demonstrate the meaning of considering statistical significances on the diameter changes of n-TiO₂ to develop more predictive QSARs.

OPEN IN VIEWER

FIG. 4.

Relationship between observed k₂ and developed k₂ from the model 1.

OPEN IN VIEWER

FIG. 5.

Relationship between observed k₂ and developed k₂ from the model 2.

In QSAR results, developed QSARs showed the good relationships between descriptors of OWCs and n-TiO₂ aggregation rates kinetics, however, the QSARs still remain unknown results on which one model is better than the other model for prediction on n-TiO₂ aggregation. However, our QSAR results push the boundaries on the general NP researches by providing some insights on the effects of OWCs on NP microscopic behavior such as aggregation in environment and moreover suggesting roles of OWCs physicochemical descriptors on NP molecular-level interaction processes in time series. Therefore, when our developed QSARs are applied to other OWCs, it should be limited to some OWCs having similar functional groups such as alkanes (R–CH₃) or hydroxyl (R–OH).