Advanced oxidation kinetics of aqueous tri alkyl phosphate flame retardants and plasticizers

Abstract

Tri alkyl phosphate esters are a class of anthropogenic organics commonly found in surface waters of Europe and North America, due to their frequent application as flame retardants, plasticizers, and solvents. Four tri alkyl phosphate esters were evaluated to determine second-order rates of reaction with ultraviolet- and ozone-generated •OH in water. In competition with nitrobenzene in UV irradiated hydrogen peroxide solutions tris(2-butoxyethyl) phosphate (TBEP) was fastest to react with •OH (k_OH,TBEP=1.03×10¹⁰ M^-1s^-1), followed sequentially by tributyl phosphate (TBP), tris(2-chloroethyl) phosphate (TCEP), and tris(2-chloroisopropyl) phosphate (TCPP) (k_OH,TBEP=6.40×10⁹, k_OH,TBEP=5.60×10⁸, & k_OH,TBEP=1.98×10 M^-1s^-1). A two-stage process was used to test the validity of the determined k_OH for TBEP and the fastest reacting halogenated alkyl phosphate, TCEP. First, •OH oxidation of TCEP and TBEP, in competition with nitrobenzene, was measured in ozonated hydrogen peroxide solutions. Applying multiple regression analysis, it was determined that the UV-H₂O₂ and O₃-H₂O₂ data sets were statistically identical for each compound. The subsequent validated k_OH were used to predict TCEP and TBEP photodegradation in neutral pH, model surface water after chemical oxidant addition and UV irradiation (up to 1000 mJ/cm²). The insignificant difference, between the predicted TBEP and TCEP photodegradation and a best-fit of the first-order exponential decay function to the observed TBEP and TCEP concentrations with increasing UV fluence, was further evidence of the validity of the determined k_OH. TBEP oxidation rates were similar in the surface waters tested. Substantial TCEP oxidation in the model surface water required a significant increase in H₂O₂.

Introduction

Recent surveys of synthetic organic compounds in wastewater receiving streams of Arkansas and Kansas found chlorinated and non-chlorinated tri alkyl phosphates (TAPs) at concentrations ranging from 0.03 to 27 μg/L (1, 2). In addition, highway runoff can be a source of TAPs in the environment as seven chlorinated and non-chlorinated alkyl-phosphates were recovered from roadside snowbanks in Sweden (3). While the eco-toxic and human health effects of aqueous TAPs are not clear, California has listed the chlorinated-TAP, tris(2-chloroethyl) phosphate (TCEP, CAS No. 115-96-8), among carcinogens and reproductive toxins since 1992 (4).

Chlorinated-TAPs have been shown to resist biodegradation, chlorination, and ozonation in treated waters (5, 6). With their resistance to conventional water treatments and relative abundance in the hydrosphere, it is apparent that alternative treatments for control of aqueous TAPs require investigation. Sampling before and after granular-activated-carbon filtration of drinking water has shown significant removal of TAPs at full-scale treatment plants (7, 8). However, the application of other advanced water treatments, specifically geared toward the chemical destruction of TAPs in-situ warrants examination.

Hydroxyl radical, with its relatively high redox potential (2.8 V), is often the oxidant of choice for engineered remediation of waters contaminated with trace anthropogenic organic compounds. The potential for •OH-formation in UV irradiated waters containing the oxidant H₂O₂, has led to the design and implementation of UV-H₂O₂ for oxidation of unwanted organics, at full-scale water treatment plants (9,10). In a previous study of •OH-oxidation of TCEP, the advanced oxidation process (AOP), O₃/H₂O₂, more efficiently degraded the aqueous TAP when compared to O₃ alone (11). The reported work assesses the potential of AOPs for remediation of surface waters containing TAPs. The advanced oxidation kinetics of TCEP were compared to a heavier molecular weight, chlorinated TAP, tris(2-chloroisopropyl) phosphate (TCPP, CAS No. 13674-84-5), and two non-halogenated TAPs, tributyl phosphate (TBP, CAS No. 126-73-8) and tris(2-butoxyethyl) phosphate (TBEP, CAS No. 78-51-3). Their second-order rates of reaction with •OH (k_OH) were assessed via a UV-H₂O₂ competition kinetics method (12). These rate constants were compared to assess the impact of alkyl-chain structure on the •OH reaction rates. Figure 1 provides a visual comparison of the structural differences among the studied TAPs.

Figure 1.

Chemical structures of TBP, TBEP, TCEP, and TCPP.

A significant application of empirically-determined k_OH is for the modeling of contaminant oxidation in natural waters via engineered, •OH-generating processes. Additionally, using previously established UV/H₂O₂-kinetic models allows for validation of empirically-determined k_OH, by comparing the ”goodness-of-fit” for k_OH-based predictive models to observed kinetic data. Therefore, for the discussed work, the derived k_OH was used to predict TAP oxidation in surface water for given solution conditions. The contaminated surface water was treated via sequential addition of a chemical oxidant, H₂O₂ or NaOCl, and a measured UV fluence (mJ/cm²). Previous work has highlighted the fast rate of •OH formation in acidic free chlorine (HOCl) solutions irradiated with monochromatic, 254 nm UV (13). TCEP and TBEP serve as model surface-water-TAPs, and to ensure an accurate prediction of their oxidation rates, the derived k_OH for these contaminants were validated using an alternative •OH-generating advanced oxidation process, O₃-H₂O₂.

Methods

Chemicals

Nitrobenzene (NB) was supplied by Fluka (Buchs, Switzerland; 97%), while TCEP, TBEP and TBP were acquired from ACROS (Geel, Belgium; 97%, 94%, and 99%, respectively). TCPP was purchased from Wako Chemicals (Osaka, Japan; purity unknown; see discussion in Analysis below). The 30% H₂O₂ solution was procured from Fisher Chemical (Pittsburgh, PA, USA). Hexachlorobenzene (HCB, >99%) was produced by Alfa Aesar (Ward Hill, MA, USA). Dichloromethane (99.9%), methanol (HPLC-grade) and chloroform (HPLC-grade with 50 ppm pentene-preservative) were supplied by Fisher Scientific (Fair Lawn, NJ). All chemicals were used as purchased. Dilute solutions were prepared in lab-grade deionized water (HYDRO Ultrapure, Research Triangle Park, NC, USA).

UV-H₂O₂ competition kinetics

The bench-scale quasi-collimated beam apparatus contained 4 low-pressure (LP) Hg UV lamps (General Electric #G15T8) emitting primarily at a wavelength of 253.7 nm. Calculating the UV fluence (mJ cm^-2) applied to an aqueous solution required a measurement of the UV irradiance (mW cm^-2) incident to the surface. A radiometer (IL 1700, SED 240/W, International Light, Peabody, MA), with a UV detector calibrated to 254 nm, was used to measure incident irradiance. Measures of incident irradiance varied between exposures due to changes in lamp condition with age and relative placement of the irradiation vessel. Therefore, the radiant exposure, or UV fluence, was averaged over the entire solution volume by multiplying an incident irradiance by factors accounting for the divergence of the collimated beam, reflection at the water surface, variation in the irradiance over the surface of the solution, and photon absorption with depth in the water column (14). In this way, multiple irradiations with variable exposure conditions can be compared.

UV-irradiated, deionized-water solutions were spiked with NB and a TAP to 5 μM (of each), H₂O₂ up to 50 mg/L, and continuously mixed. Nitrobenzene is an ideal reference compound for UV competition kinetics studies due to the compound’s resistance to direct photolysis and relatively rapid rate of reaction with •OH (12, 15, 16). The total solution volume before exposure was 130 ml, with 30 ml samples taken from the bulk solution (the change in average irradiance caused by the reduction in solution volume and subsequent decrease in time required to reach the desired UV fluence were accounted for by recalculating the required time to reach subsequent UV fluences). Each 30 ml sample was placed into a 40 ml borosilicate vial, containing 1 ml of dichloromethane, with a PTFE-lined cap. The borosilicate vials were mixed for 15 minutes on a rotary, end-over-end vial tumbler. 250 μL of dichloromethane was removed for TAP and NB concentration analysis.

O₃-H₂O₂ competition kinetics

Ozone was supplied by an EFFIZON Ozone Generator (WEDECO) utilizing compressed medical-grade oxygen. Deionized water solutions of H₂O₂ and TCEP or TBEP were dosed with ozone via addition of a measured volume of an ozone-saturated solution (40 mg/L as O₃). The molar ratio of H₂O₂ to O₃ was consistently 1.2; to observe increased TAP and NB oxidation, initial concentrations of H₂O₂ (0.38 – 6.37 mg/L) were increased with proportional increases in O₃ (0.45 – 7.5 mg/L). Ozone concentrations were measured using the Indigo-Colorimetric method as described in Standard Methods for the Examination of Water and Wastewater, 19th ed (17). After a controlled period of oxidation, sodium thiosulfate (24.8 mg/mL) was added to quench the residual O₃. A liquid-liquid extraction method, as previously described, was used to extract the analytes from the aqueous sample.

Model surface water irradiations

A model surface water was prepared by first adding extracts of Suwanee River Humic Acid (0.0256 g) and Alginic Acid (0.0532 g, International Humic Substances Society, http://www.ihss.gatech.edu) to 100 mL of 2.5 μM NaOH solution. The final “surface water” was prepared by mixing 0.09 g of NaHCO₃, 600 μL of 1 g/L NO₃^-, 2 mL of 1 g/L HPO₄^2-, and 20 mL of the buffered organic acid solution in 2 L of deionized water. Solution pH was adjusted using 0.16 N phosphoric acid (to pH 6.8). Dissolved organic carbon and inorganic anion concentrations are reported in Table 1. Significant •OH-scavengers in this surface water (CO₃^2-, HCO₃^-, and DOC) have well understood •OH scavenging rates, such that the photooxidation of TAPs in this matrix could be predicted via established equations governing the rate of •OH formation under UV/H₂O₂-conditions.

Table 1.

Concentrations of DOC and inorganic anions in the model surface water.

DOC, mg/L	NO3-, mg/L	HCO3-, mg/L	PO43-, mg/L
2.0	0.3	30	1.0

UV irradiations were conducted as previously described. TBEP, or TCEP, was added prior to irradiation (TAP₀ = 50 μg/L). Analyte extraction involved rapid vortexing of the entire irradiated solution volume (150 mL) with 1 mL of chloroform in a 250 mL centrifuge tube. A 250 μL aliquot of chloroform was taken for TAP concentration analysis.

Analysis

The analytes were separated from the solvent (dichloromethane or chloroform) with gas chromatography followed by quadropole-MS detection (Shimadzu, Japan). After injection of 1-2 μL, the column temperature was held at 50° C for 4.5 minutes; subsequent oven-temperature ramping increased the column temperature from 50 to 300° C over 13 minutes. Single-ion-monitoring was used to assess the detector response to the analytes. With regards to the purity of the TCPP standard, two peaks were detected when TCPP was the lone analyte. However, the second peak was consistently 5 times smaller (and their relative retention times were consistent), therefore only the largest peak was used for response assessment. Andresen et al. also encountered multiple peaks for their technical grade TCPP (different supplier from the TCPP used in this study) and similarly reported the largest peak for quantification (18). The mass spectra for both peaks were similar, and it was assumed that the 2^nd-order •OH reaction rates are also similar for these potential isomers.

H₂O₂ residuals were measured using the “I₃” method (19). Free chlorine residuals were analyzed according to the DPD Colorimetric method (17).

Results and Discussion

UV photosynthesized-•OH oxidation in lab-grade water

The molar absorption coefficients, ε (M^-1cm^-1), were measured for all 4 compounds by diluting the neat standard in MeOH and measuring the solution absorbance (minus the background absorbance of MeOH) over the range of 200 – 400 nm (see Supporting Information for measured molar absorption coefficients) with a Cary UV-Vis Spectrophotometer (Varian, Inc., Palo Alto, CA). Ideally, the absorbance spectra would have been measured in water rather than MeOH, due to potential differences in solute-solvent interactions. However, the poor UV-C absorbance of the studied TAPs required dissolution of solute at concentrations well above reported water solubility. Direct photolysis of TBEP, TBP, TCEP, and TCPP was insignificant at the applied UV fluences due to the almost negligible photon absorption (λ=254nm) by these alkyl-phosphates.

In solutions where NB and the tri alkyl phosphate are the only competitors for photosynthesized •OH, the linear correlation between the observed decay rate of [NB] and [TAP] represents the ratio of k_OH-NB to k_OH-alkyl-phosphate (12, 15). Figure 2 illustrates the log(ln)-linear decrease in the TAPs relative to NB in UV-H₂O₂ treated deionized water solutions of TBP & NB, TBEP & NB, TCEP & NB, and TCPP & NB.

Figure 2.

The linear-correlations between the concentrations of tri butyl phosphate (TBP), tri chloroethyl phosphate (TCEP), tri butoxyethyl phosphate (TBEP), tri chloropropyl phosphate (TCPP), and NB. •OH was the product of H₂O₂ photolysis under LP UV. Corresponding k_OH listed in units of M^-1s^-1.

The addition of a Cl-atom on the alkyl-chain (TCEP) significantly reduces H-atom availability relative to TBP and TBEP. Along with halogenation of the alkyl chain, the number of carbons and additional hetero-atoms (O) on the chain also seems to affect the rate of •OH attack. The chain structure with the most carbons (TBEP) had the fastest rate of reaction. However, TCPP has one more carbon on each chain than TCEP, yet scavenged •OH at a slower rate, most likely due to the influence of an additional function group (-CH₃) at the alpha-carbon on TCPP.

Ozone synthesized •OH oxidation in lab grade water

An additional •OH-generating process, O₃-H₂O₂, was used to measure the k_OH and validate the values determined using UV-H₂O₂. TCEP and TBEP were selected for k_OH validation as their respective rate constants are subsequently used to predict •OH oxidation in surface water (below). In ozonated aqueous solutions, the presence of H₂O₂ leads to accelerated decomposition of O₃ to •OH. As the rates of O₃-reactivity with NB (0.09 M^-1s^-1; (20)), TCEP (0.3 min^-1 in 9 mg/L O₃ solution; (11) ), and TBEP (no observed oxidation for the applied ozone Ct) are negligible relative to the reaction rate of H₂O₂ with O₃ (k_O3,HO2-=2.8±0.5×10⁶ M^-1s^-1;(21)), the assumptions underlining UV-H₂O₂ competition kinetics (no decay from UV — or ozone alone) are also valid for O₃-H₂O₂ competition kinetics. Figure 3 presents the correlations between [NB] and [TCEP] or [TBEP] in ozonated solutions of H₂O₂.

Figure 3.

The linear-correlations between the concentrations of TCEP/TBEP and NB. •OH was the product of O₃-H₂O₂.

The observed linear correlations between NB and TBEP or TCEP oxidation in ozonated H₂O₂-solutions were used to calculate k_OH,O3 for TBEP and TCEP. Figure 4 presents the experimentally determined k_OH,UV and k_OH,O3 for both TCEP and TBEP.

Figure 4.

A comparison of k_OH,UV and k_OH,O3 for TCEP and TBEP. Error bars present the upper and lower limits on the 95% confidence interval for the linear estimation of k_OH.

For both compounds, the UV-H₂O₂ competition kinetics method predicts a slightly higher rate of reaction with •OH, relative to O₃-H₂O₂. To test the homogeneity of the competition-kinetics data from UV-H₂O₂ and O₃-H₂O₂ a multivariate linear regression model

$$\ln \left(\frac{C_0}{C}\right)={\beta }_1\ln \left(\frac{{\mathit{NB}}_0}{\mathit{NB}}\right)+{\beta }_{2}z+{\beta }_3\ln \left(\frac{{\mathit{NB}}_0}{\mathit{NB}}\right):z$$ (1)

was applied to the combined UV-H₂O₂, O₃-H₂O₂ dataset. For this model, z is a dummy variable equal to 1 or 0 depending on whether the data point belongs to the UV-H₂O₂ or O₃-H₂O₂ data set. A single regression model can be fit to a combined UV-H₂O₂, O₃-H₂O₂ dataset when β₂ = β₃ = 0 (hypothesis of coincidence). Table 2 lists the parameter estimates and standard errors from the multivariate regression models for TCEP and TBEP oxidation.

Table 2.

Parameter estimates for multiple regression models (Equation 1) of TCEP and TBEP oxidation, in competition with NB for •OH.

	TCEP		TBEP
Coefficient	Estimate	Standard Error	Estimate	Standard Error
β 1	0.13	0.01	2.61	0.14
β 2	0.03	0.03	0.32	0.15
β 3	0.04	0.03	-0.17	0.38

For both TCEP and TBEP, only β₁ is significantly non-zero (α=0.05 for TCEP and α=0.01 for TBEP). Therefore a single linear regression model can be used to determine k_OH from the combined datasets of each contaminant. Table 3 summarizes the determined second-order •OH rate constants for TBEP and TCEP.

Table 3.

2^nd-order reaction rates (k_OH) of TBEP and TCEP with •OH, determined using both UV-H₂O₂ and O₃-H₂O₂ competition kinetics.

	kOH, M-1s-1
TBEP	1.03±0.38 × 1010
TCEP	5.60±0.21 × 108

TBEP and TCEP: advanced oxidation in a model surface water

The measured •OH rate constants for TBEP and TECP were utilized to predict the degradation of the contaminants under two distinct UV-based AOPs: UV-H₂O₂ and UV-HOCl (13). Theoretical first-order rate constants, k₁ (s^-1), for TCEP and TBEP degradation, in a water matrix treated with an AOP, can be predicted using established equations governing the simultaneous formation and scavenging of photosensitized •OH (22-24). The steady-state concentration of •OH ([OH]_SS) is the ratio of the rate of •OH formation, for a given set of UV-irradiation and initial oxidant conditions (H₂O₂ or HOCl), to the sum of the products of •OH-reactants and their reaction rates in the given water matrix (total •OH scavenging). The predicted k₁ for TCEP and TBEP is the product of their derived second-order •OH reaction rates and [OH]_SS. However, in UV-irradiated free chlorine solutions, the total •OH-scavenging can be significantly affected by pH; as the ratio of OCl^- to HOCl increases with pH so does the fraction of [OH]_ss that reacts with unphotolyzed free chlorine hypochlorite ion (25). Modifying the conventional equation for [OH]_ss to account for free chlorine speciation yields Equation 2.

$${\left[\mathit{OH}\right]}_{\mathit{SS}}=\frac{{\Phi }_{\mathit{HOCl}\to \mathit{OH}}{I}_0{f}_{\mathit{HOCl}}(1-e^{-2.303b({\epsilon }_{\mathit{HOCl}}{C}_{\mathit{HOCl}}+{\epsilon }_1{C}_1+\dots +{\epsilon }_n{C}_n)})}{k_{\mathit{OH},1}{C}_1+\dots +k_{\mathit{OH},\mathit{HOCl}}{C}_{\mathit{HOCl}}+k_{\mathit{OH},{\mathit{OCl}}^{-}}{C}_{{\mathit{OCl}}^{-}}+\dots +k_{\mathit{OH},n}{C}_n}$$ (2)

In Equation 2, Φ_HOCl-->OH is the quantum yield (λ=254nm) of HOCl to •OH (1.4 mol Es^-1; (13)), I₀ is the incident irradiance (Es L^-1 s^-1), f_HOCl is the fraction of UV light absorbed by HOCl, b is the optical pathlength (cm), ε_HOCl is the molar absorption coefficient for HOCl (M^-1cm^-1, λ=254nm; (13)), C₁ … C_n are the concentrations of species 1 through n (M), and k_OH,HOCl(8.46×c10⁴ M^-1s^-1; (13)) and k_OH,OCl- (8.0×10⁹ M^-1s^-1; (26)) are the second-order •OH rate constants for HOCl and OCl^-. While OCl^- photolysis products include •OH (26), consideration of OCl^- in the rate of •OH formation term (numerator in Equation 2) would not significantly impact [OH]_SS (when pH<7.5 (pKa), and particularly for water treatment applications where HOCl is the desired residual disinfectant) as its Φ_OCl->OH (λ=254 nm) is significantly less (0.278 mol Es^-1; (27)) than that of its conjugate acid HOCl.

In order to completely characterize the total •OH-scavenging for bench-scale UV-H₂O₂ or UV-HOCl treatments of surface water contaminated with either TBEP or TCEP, a model surface water with known •OH scavengers. As previously discussed, this surface water was composed of extracts of Suwanee River Humic Acid and Alginic Acid (k_OH,DOC = 2.5 × 10⁴ L/mg-C s^-1), and salts of nitrate, phosphate, and bicarbonate (k_OH,CO3 = 3.9 × 10⁸ M^-1 s^-1, k_OH,HCO3 = 8.5 × 10⁶ M^-1 s^-1) dissolved in lab-grade water (pH 6.8) (16, 28). The significant competitors for photosynthesized •OH in the prepared surface water were the spiked TAP and unphotolyzed oxidant, bicarbonate and carbonate ions, and the dissolved organic carbon. Figure 5, Figure 6, and Figure 7 provide qualitative comparisons and the degree-of-fit of the predicted to the observed TAP (TAP₀ = 50 μg/L) in model surface water treated with either UV-HOCl or UV-H₂O₂. For all cases, the predicted k₁ (s^-1) is within the confidence limits (α=0.05) of the estimate for k₁ determined by fitting a first order exponential decay function to the observed data.

Figure 5.

Observed TBEP (□) and TBEP predicted using [OH]_SS-model (—) in LP UV irradiated model surface water. Error bars present range of TBEP concentrations (2 replicates) for a given fluence. Inset table compares the predicted 1^st-order rate constant, k₁, to the estimate for k₁ determined by fitting a 1^st-order exponential decay α = function to the observed data. Upper and lower limits describe confidence interval (α = 0.05) for k₁ best-fit.

Figure 6.

Observed TBEP (□) and TBEP predicted using [OH]_SS-model (—) in LP UV irradiated model surface water. Inset table compares the predicted 1^st-order rate constant, k₁, to the estimate for k₁ determined by fitting a 1^st-order exponential decay function to the observed data. Upper and lower limits describe confidence interval (α = 0.05) for k₁ best-fit.

Figure 7.

Observed TCEP (□) and TCEP predicted using [OH]_SS-model (—) in LP UV irradiated model surface water. Error bars present range of TCEP concentrations (2 replicates) for a given fluence. Inset table compares the predicted 1^st-order rate constant, k₁, to the estimate for k₁ determined by fitting a 1^st-order exponential decay function to the observed data. Upper and lower limits describe confidence interval (α = 0.05) for k₁ best-fit.

For the given oxidation conditions, a similar reduction in TBEP (from 50 μg/L to 30, 15-18, 12, and 7 μg/L after 250, 500, 750 and 1000 mJ/cm², respectively) was observed in the model surface waters treated with UV-HOCl or UV-H₂O₂. For purposes of comparison, a UV/H₂O₂-AOP at Andijk Water Treatment Plant (The Netherlands) has been designed to remove 80% of an organic pesticide, atrazine, in lake water with 80% UV₂₅₄-transmittance (UVT), with a design UV fluence of 1000 mJ/cm² (10).

Despite yielding higher levels of •OH per mole of photons absorbed (13), the UV-HOCl process required a slightly higher fraction of the applied UV light (f_H2O2 < f_HOCl, with f determined according to previously published methods (24)) to be absorbed by the oxidant in order to achieve the same rate of TBEP oxidation as seen with UV-H₂O₂ . As previously discussed, at near-neutral pH the k_OH of the target contaminant must be similar or higher to k_OH of OCl^- (a scavenger) in order for the UV-HOCl process to be an effective contaminant barrier.

When TCEP is the target contaminant, UV-HOCl is not an effective treatment. In surface water spiked with TCEP, little to no TCEP degradation was observed in the range of applied UV fluences (up to 1000 mJ/cm²) regardless of how much NaOCl was added prior to UV irradiation. It has been previously demonstrated that at neutral pH, increasing the initial concentration of free chlorine leads to a decrease in the first-order, fluence-based rate constant for oxidation of an organic with a slower rate of reaction with •OH than OCl^- (25), as OCl^- accounts for an increasing majority of •OH-scavenging. For UV-H₂O₂ to be an effective process for oxidation of TCEP, a significant increase in the initial oxidant concentration and subsequent fraction of the applied UV light absorbed by H₂O₂ is required. As illustrated in Figure 7, a 50 percent reduction in TCEP is possible at a UV fluence of 1000 mJ/cm² in the presence of ∼50 mg/L H₂O₂. The disparate rates of reaction with •OH for the TAPs, competitive radical scavenging by natural organic matter, carbonate species, and unphotolyzed oxidant concentrations are all parameters that impact the selection of chemical oxidant to be photolyzed, and subsequently, the TAP-removal-efficiency with UV-AOP.