Studies of Emission Processes of Polymer Additives into Water Using Quartz Crystal Microbalance-A Case Study on Organophosphate Esters.

Plastic materials contain various additives, which can be released during the entire lifespan of plastics and pose a threat to the environment and human health. Despite our knowledge on leakage of additives from products, accurate and rapid approaches to study emission kinetics are largely lacking, in particular, methodologies that can provide in-depth understanding of polymer/additive interactions. Here, we report on a novel approach using quartz crystal microbalance (QCM) to measure emissions of additives to water from polymer films spin-coated on quartz crystals. The methodology, being accurate and reproducible with a standard error of ±2.4%, was applied to a range of organophosphate esters (OPEs) and polymers with varying physicochemical properties. The release of most OPEs reached an apparent steady-state within 10 h. The release curves for the studied OPEs could be fitted using a Weibull model, which shows that the release is a two-phase process with an initial fast phase driven by partitioning of OPEs readily available at or close to the polymer film surface, and a slower phase dominated by diffusion in the polymer. The kinetics of the first emission phase was mainly correlated with the hydrophobicity of the OPEs, whereas the diffusion phase was weakly correlated with molecular size. The developed QCM-based method for assessing and studying release of organic chemicals from a polymeric matrix is well suited for rapid screening of additives in efforts to identify more sustainable replacement polymer additives with lower emission potential.

Introduction

Plastics are widely used both in industry and in our daily life, which has caused a rapid increase in production to a global production rate reaching 348 million metric tons in 2017.[1] To improve the performance and functionality of the final products, most virgin polymers are first compounded with additives and then manufactured with processing aids, which are usually not covalently bound to the polymer.[2] As a consequence, release of ancillary chemicals from plastic products occurs in all phases of the product life cycle,[3−7] posing risks to the environment and human health.[8] It has been reported that the rate and extent of release of additives can be affected both by their properties (e.g., water solubility, hydrophobicity, and molecular weight) and those of the polymers (e.g., composition, glass transition temperature (Tg), and crystallinity), as well as by environmental factors, such as temperature and weathering processes.[5,9] Current research on plastic pollution focuses mainly on effects and fate of microplastics or their potency to accumulate and transport hydrophobic organic chemicals. Only a few studies have been conducted on release of additives from plastics to water, despite the high contents of additives in plastics and the adverse health effect of many of these compounds.[10−12] Plastic waste in contact with water is a significant source of environmental contamination by plastic additives, as plastics are used in plumbing and drainage systems, medical devices, and for outdoor use such as furniture and tarpaulin.[13−15] It is therefore of great importance to increase our understanding of the release processes of additives from plastics to the aquatic environment and to find new chemicals with lower emission potential.

Recently, Sun et al. has studied the release kinetics into water of polybrominated diphenyl ethers (PBDEs) and 1,2-bis(2,4,6-tribromophenoxy)ethane from microplastic pellets made of acrylonitrile butadiene styrene terpolymer. They could predict the diffusion coefficients of brominated flame retardants (BFRs) in other types of microplastics based on the observed semiempirical linear relationship between the logarithm of diffusion coefficient (log D) of PBDEs and the Tg of plastics.[12] Paluselli et al. reported the release kinetics of phthalates from the polyvinyl chloride (PVC) cable and polyethylene bags into seawater under varying light exposure, bacterial density, and temperature.[5] In most cases, these release experiments were performed in the field by submerging plastic debris in seawater,[16] in laboratory settings by analyzing concentrations of additives in the aqueous solution continuously,[5,17] or by comparing the gravimetrical changes of the plastics before and after water exposure.[18] Although these experimental conditions are closer to those in the environment, these methods usually take at least few weeks or even hundreds of days to collect enough data. Long experimental periods may increase the risk for losses by adsorption, evaporation, (bio)transformation reactions, or changes in diffusivity because of formation of biofilms on the plastic surfaces,[19] which increases the uncertainty of these methods. In addition, the limited number of data points produced could make it difficult to discriminate between different transport models and thus increase the uncertainty in interpretation of modeling results and further prediction. Therefore, it is urgent to develop more rapid and accurate methodologies.

Quartz crystal microbalance (QCM) is a high-accuracy technique that can record real-time mass changes of coatings applied to the surface of piezoelectric crystals in contact with liquids at the nanogram scale and has been extensively used in characterizing biomolecular binding events at solid/liquid interfaces.[20] The technique was recently applied to investigate the deposition and release kinetics of nanomaterials on/from silica surfaces, as well as the deposition/desorption of low volatility toxins on microplastics in situ and in real time.[21−23] It has also been used to measure the deposition of phthalates from the gas phase on QCM crystal surfaces coated with silicon using an electron beam technique.[24] However, QCM has not been applied to study release of polymer additives to water.

Organophosphate esters (OPEs) are a group of plastic additives, which are produced in large volumes and used extensively in commercial products as plasticizers and flame-retardants to replace BFRs.[25−27] Certain OPEs are known or suspected carcinogens or neurotoxic substances,[6] and several studies have reported on their occurrence and distribution in indoor air, dust, and in aquatic environments.[7,28] Liang et al. recently determined the parameters controlling gas phase emissions of OPEs in the indoor environment.[4] However, the release kinetics of OPEs from polymers into the aquatic environment remains unclear, as does the variations of OPE release patterns in relation to polymer properties and environmental factors.

In this study, we used various OPEs as model compounds to develop a novel approach based on QCM for measuring time-dependent release from three different polymer films to the water phase in real time. Empirical models were fitted to the mass changes detected by the QCM to reach a better understanding of its relationship to the physicochemical properties of the studied chemicals. In addition, Hansen solubility parameters (HSPs) of the studied OPEs were predicted as a basis for calculating the HSP distance between OPEs and polymers (DHSP), aiming at further understanding the relationship between the solubility of additives in polymers and their release kinetics. The QCM-based approach together with the derived parameters provide a basis for quantifying and predicting the release potential of polymer additives and thus assessing risks for environmental and human exposure of emerging contaminants.

Materials and Methods

Materials

Polystyrene (PS) and poly(methyl methacrylate) (PMMA) were purchased from Sigma-Aldrich; PVC was obtained from KEBO Lab (Stockholm, Sweden). The OPEs used in this study were tris(2-chloroethyl)phosphate (TCEP; 97% purity), tris(2-butoxyethyl)phosphate (TBEP; 94% purity), tributylphosphate (TBP; 99%), and triphenyl phosphate (TPP; 99%) obtained from Sigma-Aldrich (St. Louis, MO, USA); tris(1-chloro-2-propyl)phosphate (TCPP) and tris(1,3-dichloro-2-propyl)phosphate (TDCPP), both of technical quality, purchased from Albemarle (Charlotte, NC, USA); 2-ethylhexyl diphenyl phosphate (EHDPP; 96%) purchased from Chiron AS (Trondheim, Norway); (3-diphenoxyphosphoryloxyphenyl)diphenyl phosphate (RDP; technical quality) as Fyrolflex RDP purchased from AkzoNobel (Arnhem, The Netherlands); and tris[3-bromo-2,2-bis(bromomethyl)propyl]phosphate (TTBNPP; 98%) purchased from Combi-Blocks (San Diego, CA, USA). The structures and physicochemical properties for the polymers and OPEs are listed in the Supporting Information (Tables S1 and S2). Toluene, chloroform, and tetrahydrofuran (THF), all of HPLC grade, were purchased from VWR (Stockholm, Sweden). Ultrapure water was produced by a Milli-Q Advantage Ultrapure Water purification system (Millipore, Billerica, MA, USA) and filtered through a 0.22 μm Millipak Express membrane.

Preparation and Characterization of Polymer/OPE Films

Polymer films containing OPEs were prepared by spin-coating, following reported procedures with some modification.[23,29] In brief, polymer solutions were obtained by mild sonication in appropriate solvents in a water bath for 30 min (50 mg mL–1 PS solutions in toluene, 30 mg mL–1 PVC solutions in THF, and 30 mg mL–1 PMMA solutions in chloroform). Corresponding amounts of OPEs were weighed and added to separate polymer solution aliquots and then mixed together. Then, polymer/OPE films on gold-coated AT-cut piezoelectric quartz crystal sensors (Novaetech S. r. l., Italy) were obtained by spin-coating at 3500 rpm for 30 s. All films were prepared 1 day before the QCM measurement. A similar spin-coating procedure was used to prepare films of pure polymer or polymer/OPE mixtures on glass substrates for surface morphology characterization and for measurements of contact angle and film thickness. Scanning electron microscopy (SEM) (Merlin FESEM, Carl Zeiss, Germany) was used to investigate the surface morphology of polymer/OPE films. Static contact angle measurements were obtained using a θ contact angle meter from Biolin Scientific (Göteborg, Sweden). Film thickness was measured using a DektakXT stylus profilometer (Bruker, Billerica, MA, USA). Further details on spin-coating procedures and characterization of polymer/OPE films are given in the Supporting Information.

QCM Measurements and Analysis

The release patterns of the tested OPEs from the corresponding polymer films were recorded at room temperature (22 °C) using an openQCM Wi2 QCM instrument (Novaetech S. r. l., Italy), which has a sensitivity of ∼4.42 ng Hz–1 cm–2 and a detection limit of 1.25 ng in liquid (Figure S1). To investigate the temperature effect on release patterns of OPEs, the water container and QCM fluidic cell were placed into a gas chromatography (GC) oven with a stable and accurate temperature. The flow rate was 3.0 mL min–1 in all experiments, except those designed to investigate the influence of flow rate. Release of OPEs from polymer films led to a decrease in the mass loading of the quartz crystal, thereby increasing its resonance frequency. Considering that the polymer films were dried prior to testing and the inability of water to swell the chosen polymers (corroborated by the absence of apparent increases of mass in any of the experiments), it is reasonable to assume that the polymer films coated onto the quartz crystal are rigid, and thus the resonance frequency changes, Δf, are directly proportional to the mass changes (Δm) on the quartz crystal, according to the Sauerbrey relationship[30]where f0 is the resonant frequency of an uncoated quartz crystal (10 MHz), using A = 0.2826 cm–2 for the active, gold-coated electrode-covered area of the quartz crystal, ρq = 2.648 g cm–3 for the density of quartz, and μq = 2.947 × 1011 g cm–1 s–2 for the shear modulus of quartz crystal.

Calculation of HSPs

HSP is a combination of the three components δD, δP, and δH (explained in Supporting Information). HSP values for studied polymers were taken from the literature,[31−33] and an average of each HSP component was calculated (Table S1). The approach to derive HSP data for each of the studied OPEs, which was not available in the literature, is described in detail in the Supporting Information. In brief, models were developed using partial least squares regression[34] with HSP data for 71 similar compounds from the literature[35] (see Supporting Information) and a set of 65 calculated molecular descriptors (MOE[36]) that had been used in previous studies.[27,37,38] These models were applied to predict HSP parameters for the studied OPEs (Table S2). DHSP was then calculated to provide a rough estimate of the solubility of each OPE in the studied polymer films following the equation[39,40]where δD1, δP1, and δH1 are the HSP values for polymers, and δD2, δP2, and δH2 are the HSP values for OPEs. Calculated DHSP is listed in Table S2.

Quality Assurance and Quality Control

The repeatability of the QCM approach was assessed by comparing the variations in six repeated release curves from 30 wt % TPP in PS films. Furthermore, the initial content of OPEs in polymer/OPE films after spin-coating was also measured by GC–mass spectrometry (GC–MS) using an Agilent 5975 instrument (Agilent Technologies, Palo Alto, CA). After spin-coating, three quartz crystals coated with PS films containing 30 wt % of TCEP were extracted three times with 8 mL aliquots of methanol, which were spiked with 2H-labeled TCEP as an internal standard. The combined extracts were concentrated to 1–2 mL by rotary evaporation and further concentrated to 1 mL by nitrogen blowdown. A ZB-5MS capillary column (60 m, 0.25 mm ID, 0.25 μm film thickness; Phenomenex, CA) was used for GC–MS analysis. The GC oven temperature program was as follows: initial temperature was set at 90 °C (held for 2 min), increased to 190 °C at 15 °C min–1, and then increased at 5 °C min–1 to a final temperature of 300 °C. TCEP was monitored at m/z 249 and analyzed with GC–MS.

Results and Discussion

Development of the QCM-Based Method to Study Emissions of Polymer Additives

Emissions from PS films with 20, 30, 40, and 50 wt % initial concentrations of TCEP and TPP were measured, and the initial release from the PS films prepared with 30, 40, and 50 wt % TCEP was very fast (Figure a). These levels were chosen to resemble use of OPEs as a plasticizer (10–70 wt %) or a flame-retardant (3–25 wt %).[2,41] Emissions of pure PS film were measured as a reference (Figure S6). During the first second of the emission phase, the frequencies of the crystals coated with PS containing 40 and 50 wt % of TCEP had already increased to levels corresponding to the release of 13 and 20 wt % of TCEP, respectively. This indicates that syneresis had taken place, causing a significant fraction of the TCEP to be present on the surface of the PS films with the highest loading (40–50 wt %). Following the very fast initial mass loss, the release rates gradually slowed down and approached plateaus after about 1 h. In contrast, the release rate of the PS film prepared with 20 wt % of TCEP was much slower at the beginning. Similar release patterns were observed for PS films containing 30, 40, and 50 wt % of TPP, which showed faster release at the initial stages than the PS film with 20 wt % of TPP (Figure b). This could also be attributed to the increasing free volumes of PS caused by a plasticizing effect of OPEs at higher concentrations and thus faster migration and diffusion rates of OPEs.[42]

Figure 1

Release patterns of (a) TCEP and (b) TPP from PS films containing 20, 30, 40, and 50 wt % of TCEP and TPP. For y axis, m0 means the initial OPE mass, and m means the OPE mass at time t.

We observed porous reticulated structures and characteristic of phase separation, using SEM for PS films containing 30 wt % of TCEP and above; the 50% TCEP sample also showed holes in the ridges, where trapped TCEP may have escaped to the surface during film annealing (Figure a–d). This segregation of the polymer phase cannot be caused merely by surface dewetting[43] because the film with 20 wt % of TCEP exhibited a smooth and featureless surface, similar to that of the PS film prepared without OPE addition (Figure S7a). Besides higher concentration gradient of TCEP between polymer films and water phase, spontaneous demixing, where a large fraction of the TCEP is expelled to the surface at the drying stage, could hence have contributed to the fast initial TCEP release rates. Alternatively, the TCEP molecules inside the polymer film may have transferred quickly to the polymer–water interface through interconnected TCEP channels formed because of phase separation between polymer and TCEP at higher concentrations.[44,45] We also tested PS films containing 10 wt % of TCEP, but no significant release of TCEP could be detected over 3 h by our QCM system. This indicates that TCEP was not capable of forming a continuous, interconnected biphasic system at low concentrations, preventing its rapid release from polymer films, or, alternatively, as one reviewer noted, the low release rate could be explained by a low concentration gradient of TCEP in the PS film and at its water interface, yielding emissions of TCEP below the detection range of the QCM. Similarly, phase separation was also observed in the PS films prepared with TPP at 40 and 50 wt %, which showed bicontinuous wavy structures in the SEM images and characteristic of spinodal decomposition (Figure e,f).[46] These undulated structures could lead to a somewhat higher specific surface area, which could at least partly explain the higher release rates at higher TPP content as compared with lower levels (20 and 30 wt %).

Figure 2

PS films characterized using SEM containing (a) 50 wt % of TCEP, (b) 40 wt % of TCEP, (c) 30 wt % of TCEP, (d) 20 wt % of TCEP, (e) 50 wt % of TPP, (f) 40 wt % of TPP, (g) 30 wt % of TPP, and (h) 20 wt % of TPP.

TCEP was released at higher rates than TPP independent on additive levels in the polymer. TPP was however released in larger amounts than TCEP when blended in PS at 40 and 50 wt %. The slower release rate of TPP is well in line with a higher log KOW value, lower water solubility, bulkier molecule, and the aromaticity of TPP, which yields lower partitioning to water at the polymer–water interface and thus a higher polymer–water partition coefficient. In addition, TPP can engage in π–π interactions with PS and may exhibit greater steric hindrance in PS films caused by the aromatic groups in TPP as compared with TCEP, which could reduce the diffusion rate of TPP in PS films.[12] In this study, the experimental period to reach the apparent steady state was significantly shortened, compared with previous studies covering periods over weeks or even months using plastic debris from real products.[12,16,17,47,48] This is because thin polymer films (427 ± 67 nm) were used in this study (Figure S8), which significantly increase their specific surface area and shorten the diffusion paths of OPEs from the bulk polymer to the polymer–water interface.[5] Previous studies have reported significantly increased release rates of additives from particulate plastic matrices with reduction in the particle size and thus by shorter diffusion paths and an increase in their specific surface area.[12,16] Very low emission rates have been reported from compounded plastic pellets, which were obtained by crushing commercial products to sizes from few micrometers to millimeters.[12,47,48] These compounded plastic pellets are chemically more complex than the films prepared from pristine polymers by the spin-coating method used in this study, as they contain numerous chemical additives that could affect the material characteristics and likely influence the emission rates of each additive.

Considering the effects of the aforementioned morphological features of the polymer films on the OPE release patterns, experiments were conducted using polymer films with 30 wt % of OPEs (i.e., where no variation in morphology was observed at the magnification level of the SEM images). The repeatability of the developed method for TPP release was assessed by measuring six individual PS films containing 30 wt % of TPP, which resulted in a standard error of ±2.4% (95% confidence interval) at the end of the plateau (see also Figure S9). In addition, the accuracy of the developed QCM method was validated by analyzing the actual content of TCEP in PS films with 30 wt % of TCEP using a GC–MS-based method. In the analysis, we assumed that the mass fraction of TCEP in the PS films coated on quartz crystals was identical to its original fraction in the polymer/TCEP solution and that no losses had occurred prior to analysis. A paired t-test revealed that there was no statistically significant difference in the measured TCEP levels in the GC–MS- and QCM-based analyses (Table S3), and a two-sided F-test verified that the variances for the two methods were significantly different reflecting variation in sample preparation and analysis. This analysis confirmed that the TCEP content in the polymer film after spin-coating was in good agreement with the desired concentration (30 wt %) and that the QCM method resulted in data with low variation. Under the conditions tested in the current study, the flow rate showed low impact on the release rates in the range of 1–4 mL min–1 using 30 wt % of TPP in PS (Figure S10).

Release of OPEs from PS Films

Nine OPEs were selected for comparison of their release patterns from PS films and these include TCEP, TCPP, TBEP, TDCPP, TBP, TPP, EHDPP, RDP, and TTBNPP. The compounds were selected to cover a large range in hydrophobicity, that is, from 1.6 to 8.0 in log KOW and different structural functionalities including various halogens, alkyl-chains, and aromatic structures (Table S2). The plot of the release of these OPEs at 30 wt % in PS films with initial masses of ca. 13 μg (Figure and Table S4) indicated that the emissions of all OPEs reached the apparent steady state within 10 h, except for EHDPP (∼32 h) (Figure S11). It was noted that each of the PS/OPE films studied exhibited smooth and featureless surface without obvious phase separation characteristics (Figure S12b–i), except PS films with 30 wt % of TCEP (Figure S12a), indicating that the differences in release kinetics of the OPEs originate mainly from their variation in physicochemical properties and interactions with the polymer matrix. Hydrophobicity and molecular size of additives have in previous studies been attributed as the cause for variation in release patterns.[49−53] Most OPEs exhibited large releases from the PS films, except RDP and TTBNPP, which released only less than 5% of their original content in 10 h. The slow release of these two compounds from PS could be because of their low water solubility, evident from high log KOW values, but also because of their bulky molecular structures, yielding large steric hindrance in the polymeric matrix and thus decreased diffusion rates.[9] Although EHDPP has a comparable log KOW value and a similarly bulky molecular structure as RDP and TTBNPP, the release of EHDPP was much higher (40% after 10 h, Figure ). This could be explained by the relatively high water solubility of EHDPP, which is 600 and 2 × 106 times higher than those of RDP and TTBNPP, respectively.

Figure 3

Release patterns of tested OPEs at 30 wt % loading in PS films.

Numerous empirical models have been used to describe the mass transfer processes between two media,[54,55] including release of organic compounds from polymers into, for example, air, water, or biota tissues.[56−58] Attempts to fit the experimental data to first-order (linear) or second-order (single-exponential) emission models did not succeed, indicating that the release processes are driven by more than a single factor (e.g., dissolution, partition, or diffusion). Several different empirical models fitted well with the measured data (Table S5), however, as discussed by Wells et al. many of these empirical models tend to be overparameterized and may lead to faulty conclusions when applied to the emission mechanisms.[58] For example, multiple parameter exponential models are commonly used to describe additive emissions from microplastics or polymer products[54,59] and were capable of fitting the data from the present study very well (Table S5). However, a sensitivity analysis of these models indicated overfitting, likely because of the large number of independent parameters (three parameters for the double exponential and five for the triple exponential). Another kind of widely used modeling approach is the mass balance-based semiempirical models.[12,61] However, as recently discussed by Wells et al.,[58] although these models are more fundamental compared to purely empirical models, they are frequently overparameterized and requires a lot of assumptions that are not always true, making the model quality hard to be determined from fitting results. For example, one of the assumptions shared by these models is a constant diffusion coefficient, which is not true for our experimental setup. Therefore, the one-compartment Weibull distribution model[58,62] was tested because it has only two fitting parameters and thus appears to have lower risk of overfitting[58]where m0 is the initial weight of OPEs, and m is the weight of OPEs at time t. The Weibull scale parameter α and shape parameter β were obtained by iterative curve fitting, which resulted in fits with correlation coefficients varying from 0.90 to 0.99 (Table S6).

Sensitivity analysis of the Weibull parameters was conducted by varying the α (log transferred) and β values up to 25%. The results showed the risk of overfitting to be much lower for the Weibull model compared to the other tested empirical models. The model should be interpreted by the parameter α having an impact on the initial fast releasing period, whereas β is related to the second and slower part of the release phase. This agrees with previous studies showing that the release process has a fast linear releasing phase dominated by partitioning of additives between the polymer surface and the water phase,[61] followed by a slow phase dominated by diffusion in the polymer.[9,51,56,63] For the studied OPEs, log α was found to be linearly correlated to the water solubility of the compounds (log SW, R2 = 0.77, Figure S13a) and their hydrophobicity (log KOW, R2 = 0.95, Figure S13b). Previous studies have also shown that the polymer–water partition coefficient is correlated to water solubility[65] or log KOW.[49−51,65,66] The partition phase has also proven to be dependent on the film thickness and concentration gradient between the plastic and water.[61] The influence of concentration gradient is easily seen in Figure , while the impact of film thickness remains to be studied, as the films used in this study have similar thickness. As for the diffusion-related parameter β, RDP and EHDPP showed higher β values (0.62 and 0.63) while the other eight OPEs show relatively similar values (0.15–0.33). The diffusion coefficient is typically correlated with molecular size of polymer/additives or strength of chemical interactions between the polymer and additive.[9,67−69] For the nine studied OPEs, the β parameter showed a slightly better correlation with molecular volume (R2 = 0.37, Figure S13c) than with the solubilities of OPEs in PS reflected by the predicted DHSP values (R2 = 0.30, Figure S13d). Note here that the applied DHSP data are derived from a QSPR model trained on a combination of experimental and estimated data, and the predicted DHSP values for the nine OPEs showed a very low variation (Table S2). The uncertainty in the DHSP data might thus have a high impact. The highest β value of RDP and EHDPP cannot be explained by its molecular size (molecular weight or volume), hydrophobicity (log KOW), or its solubility in PS (DHSP, Table S2), indicating that the slower release process, likely diffusion-hindered, is not directly controlled by any of these single features. Other factors influencing the diffusion may be involved, such as the chemical structure of the OPEs, and for example, the high β value of RDP may have been caused by the five aromatic rings. A previous study showed that besides molecular size, other polymer-related parameters, for example, Tg and aging of the polymers, can have an impact on the diffusion.[61] The rate-limiting diffusion-controlled phase is the most critical for estimating release to the environment and for the eight studied OPEs emitted from PS (EHDPP excluded), and the diffusion processes between 300 and 600 min can be considered linear with similar release rates (0.3–0.8% of the initial mass per hour, Figure S14). In the initial phase of the curves, a linear and rapid drop can be observed for most of the OPEs, indicating that there might be surface wash-off during the first few minutes. However, a modeling effort of the release where the first 5 min were omitted resulted in nearly similar Weibull coefficients, indicating a two-phase emission process with the first phase controlled by partitioning.

Effect of polymer/OPE Interactions on Release Kinetics of OPEs from Different Polymers

The release of a selected set of OPEs (TBP, TDCPP, and TPP) was measured from the polymers PS, PVC, and PMMA to study variation in polymer/OPE interactions (Figure ), with emissions of pure PS, PVC, and PMMA films as references (Figure S6). TBP, TDCPP, and TPP were selected because they have similar log KOW values but differ in the nature of the substituents; linear butyl groups (TBP), chlorinated isopropyl groups (TDCPP), and phenyl groups (TPP). The polymers PS, PVC, and PMMA differ in their physicochemical properties. For example, the measured contact angles, which reflect hydrophobicity of polymer films, varied from 89.3 ± 0.8° for PS to 86.7 ± 4.3° for PVC and 63.6 ± 5.0° for PMMA (Table S1). Their glass transition temperatures (Tg) vary from 105 °C for PS and PMMA to 70 °C for PVC, and the HSPs, δH, corresponding to the energy in hydrogen bonding between the additive and polymer, differ from each other (i.e., 4.3, 5.8, and 7.5 for PS, PVC, and PMMA, respectively) (Table S1). PMMA exhibited the lowest release among the studied polymers at 30% loading, with TBP releasing 10%, TPP releasing 23%, and TDCPP releasing 2%, in 10 h. This is likely because of more rigid PMMA segments and lower free volume than PVC, which could decrease the diffusivity of OPEs, and thus lead to lower release from PMMA.[12] These characteristics are also reflected in higher Tg value of PMMA compared to PVC, even though the addition of OPEs may decrease Tg values of polymers to some extent. PS has a Tg similar to PMMA (Table S1) but lower δH indicating lower hydrogen bonding capacity and it may thus be less rigid, resulting in higher emissions. PMMA has also slightly longer chain length on average compared with PS and PVC, according to their molecular weight (Table S1), indicating decreased flexibility of the chain backbone and thus potential increased steric hindrance, yielding reduced polymer diffusion.[70]

Figure 4

(a) Molecular structures of TBP, TPP, TDCPP, PMMA, PS, and PVC; (b) release patterns of TBP, (c) TPP, and (d) TDCPP at 30 wt % in PS, PVC, and PMMA films, respectively.

TDCPP showed lower emissions after 10 h from the three polymer films compared with TBP and TPP with the exception of TBP in PS showing a similar released amount. TDCPP has a lower total emission likely because of heavier and bulkier substituents and thus larger molecular volume and weight, as compared with TBP and TPP (Table S2). The variation in release from the different polymers cannot be explained by DHSP because the Hansen parameters of all the tested polymers are similar (Table S1). This indicates that these parameters are not able to describe the variation in release between these three commodity polymers, and that there are other factors driving these variations. In addition, these polymer/OPE films also exhibited a smooth and featureless surface as other PS/OPE films (Figure S15).

The release data of TDCPP, TBP, and TPP from the three polymers were also fitted using the Weibull model (see above) (Table S7). A low variation in the β coefficient was observed, which implies that variation in diffusion was not the major driver of the observed differences in released amounts. However, the α value showed a larger variation, in particular for PVC and PMMA, indicating the importance of the initial partitioning phase. Although both α and β were found to correlate with chemical properties, such as hydrophobicity and molecular volume, for the eight OPEs released from PS films, similar correlations were not observed for the three tested OPEs released from PVC or PS, neither with other properties, for example, water solubility, DHSP, nor molecular weight. This indicates that it is not reliable to use a single property to estimate both the partitioning and diffusion processes, especially within a small set of chemicals with limited variations in physicochemical properties.

OPE Release under Varying Temperatures

The release profiles of TPP from PS and PVC films were measured at 22, 30, and 40 °C (Figure S16). As expected, the amount of TPP emitted over 200 min from both the PS and PVC films increased with rising temperature, as did the initial release rate of TPP. The amount of TPP emitted from PS increased by 74 and 130% at 30 and 40 °C, respectively, as compared with 22 °C. For PVC films, the emitted amounts increased even more, that is, by 87 and 186% at 30 and 40 °C, respectively. The initial release rates of TPP from both PVC and PS in the first 10 min increased by factors of 10 and two for PVC and PS films as the temperature was increased from 22 to 40 °C. Generally, the addition of additives acts to weaken intermolecular interactions between polymer chains and thus decreases the Tg of polymers. More additives can be released from the polymeric matrix at higher temperatures because of faster molecular diffusion and higher flexibility of the polymer chains.[9] Our results are in general in agreement with previous studies, for example, release of BFRs and OPEs from plastics has been seen to accelerate significantly with temperature.[71] Waye et al. reported that the emission rate of BDE-209 from a computer case increased by 75–80% for each 5 °C increase in temperature (between 30 and 45 °C), and similar trends were shown for other congeners.[72] The difference in release amount of TPP between PVC and PS may be attributed to variations in polymer characteristics. Because PVC has lower Tg (70 °C) than PS (105 °C), the segment mobility within the PVC chains is higher. As a consequence, it has a less integrated structure and larger free volume between PVC chains. The larger free volume could result in a decreasing resistance for diffusion of the molecules in the polymer and thus higher release of TPP from PVC films. This agrees well with a previous study showing that polymers with low Tg values have greater chain segmental mobility, allowing diffusive molecules to move more easily.[12] PS, PVC, and PMMA used in current study are all atactic, with 18 and 6% isotactic triad tacticity for PVC and PMMA, respectively (Figure S17), which might affect the Tg values of the polymers. We are also aware of that slightly lower Tg values usually occur for polymer films with thickness from several microns down to a few hundred nanometers than for bulk polymers.[73]

Environmental Implications

The present study presents a novel approach to measure release of additives to water using QCM from various types of thin polymer films. The method generates reproducible and fast real-time release data with a standard error of ±2.4% (at 95% confidence interval), and apparent steady-state transports were reached on an hour scale. It allows for studying influence on additive emissions of aquatic environmental factors, including variation in pH, salinity, temperature, and dissolved organic matter in the water. Another option would be to simulate microplastic properties by, for example, weathering the plastic films including photodegradation or by formation of biofilms or other environmentally relevant processes of concern. QCM has previously been successfully used as a mass sensor for vapor sorption/desorption in the gas phase.[74,75] Therefore, we feel confident that the presented QCM approach can be extended to studies of emissions of organic compounds from polymer matrices into the gas phase. The QCM approach is, however, not able to differentiate between emissions of several additives simultaneously. It is also limited to lab-made polymer films, and the measured emission rates are not directly representative of a real situation. Actual products usually exhibit much lower specific surface area and more intact structures than the lab-made polymer films, which result in lower emissions. Further studies are needed to apply the QCM approach to actual plastic products.

Although plastic products usually contain high contents of OPEs, our results indicate that more than 60% of OPEs or even 98% of extremely hydrophobic OPEs remain in the polymer and will be slowly released for an extremely long period of time. However, levels of OPEs emitted from polymers increase with rising temperature. The temperature-dependent emission process should be considered in particular considering global warming and its potential impact on additive emissions from plastics in landfills and debris at sea. Although the presented data does not account for the complexity in size, morphology, and composition of real plastic products, as well as weathering effects, this study provides insights into the release process and variation in release of OPEs from plastics on a molecular level. Emissions of additives from plastics represent a significant potential source of emerging contaminants and should be considered in environmental and human health risk assessments. The presented QCM-based method was capable of determining differences in release characteristics of OPEs with their physicochemical properties and the characteristics of the polymer. It therefore provides an alternative screening approach to derive data for gaining better understanding of additive emissions from polymeric materials in general. As such, it has a potential to be used for ranking of plastic additives in attempts to identify more sustainable candidates with lower emission rates.