The newly developed quantification method is based on complete methanolysis of the polyester PET to DMT using CH3ONa as basic catalyst (see Fig. 2). The formed DMT is determined by GC–MS.

Fig. 2

Reaction equation of methanolysis of PET

Method optimization and validationOptimization of methanolysis

For optimization of the methanolysis step, different catalysts, reagent amounts, and reaction times were tested. Pham and Cho [14] described methanolysis of PET in an equimolar mixture of DCM and MeOH using K2CO3 as catalyst at room temperature. This method was used as a starting point for optimization. While yields for DMT were sufficient (90–100%) at a PET concentration of 1 mg g−1, yields at smaller concentrations (< 100 µg g−1) were not satisfactory (40–70%). This was caused by the formation of terephthalic acid or the monoester by hydrolysis, which was also reflected in a decrease in IS intensity.

Water-free conditions were tested to avoid this side reaction. However, without water, no depolymerization of PET took place. Apparently, a certain amount of water is needed for the activation of the catalyst K2CO3. However, this challenge was solved by using sodium methoxide (CH3ONa) as catalyst as described by Cao et al. [15]. CH3ONa is soluble in the used solvent mixture (MeOH/DCM), which should be water-free. By this approach, recoveries > 90% were attained over a broad concentration range (0.005–1 mg g−1).

Since natural matrix compounds can react with the catalyst, they can hamper the efficiency of the chemolysis. In order to optimize the amount of sodium methoxide needed, recovery experiments were carried out in an authentic environmental matrix (SPM). Quantitative yields of 90–100% DMT were obtained using 100 µL CH3ONa (0.5 M in MeOH), whereas 50 µL only led to conversion rates of 30–71% (data in SI, Table S1). In order to optimize the overall reaction time, the conversion rate of PET to DMT was determined after 2, 4, 8, 16, 24, and 48 h of methanolysis (see Figure S3 in SI). It was found that methanolysis was quantitative after 16 h. Due to a convenient time-scaling in the laboratory, a period of 24 h was chosen as the final reaction time.

Calibration

The preparation of different calibration points from a stock solution by serial dilution was impossible, since PET is only soluble in few solvents such as 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP), and even traces of this alcohol lead to byproducts such as methyl 1,1,1,3,3,3-hexafluoro-2-propanyl terephthalate and bis(1,1,1,3,3,3-hexafluoro-2-propanyl) terephthalate if K2CO3 was used as catalyst (ref. Figure S4). No chemolysis of PET was observed in the presence of NaOCH3 and HFIP. This can be explained by higher acidity of HFIP compared to MeOH and lower nucleophilicity of the anion caused by the electron-attracting CF3-groups. Direct weighing of the polymers was also an unsuitable method, due to uncertainties of weighing of small quantities when preparing the lower calibration levels. Therefore, as an alternative, a serial dilution of the stock mixture was carried out with calcined sea sand, an inert matrix. This approach has already been shown to work for other MP types (PE, PP, PS) [1]. Thus, calibration points were generated by analyzing known concentrations of PET diluted with sand. A sand/PET-d4 mixture was added as IS before start of the methanolysis. During method development, no deuterium/hydrogen exchange was observed. The deuterium labeling at the benzene ring was stable under the conditions of methanolysis. For the calibration curves, relative peak areas were calculated with regard to DMT-d4. A linear calibration curve over a range of 0.05–50 µg PET was achieved, which equates to concentrations of 1–1000 µg g−1 for a sample weight of 50 mg (see Figure S5 in SI).

Limits of detection and quantification

The limit of detection (LOD) and limit of quantification (LOQ) were derived from the signal-to-noise ratio (SNR) in diluted PET standards (1–10 µg g−1, i.e., 50–500 ng absolute PET mass), with an SNR ≥ 3 required for the LOD and an SNR ≥ 10 required for the LOQ. An LOD of 50 ng PET and an LOQ of 200 ng PET were achieved. With a typical sample in-weight of 50 mg, this equates to concentrations of 1 and 4 µg g−1, respectively. For a volume of 9 L water, this equates to concentrations of 6 and 22 ng/L, respectively.

Matrix interferences

Quantification of PET is based on the indirect determination of the polymer by its methanolysis product DMT. However, DMT might also be formed from matrix ingredients present in environmental samples. Therefore, the selectivity of the method was evaluated by analyzing organic materials which were not contaminated by PET. Table 2 provides the bias for the various natural compounds. In general, the tested matrices led to a negligible bias above the LOQ, except for humic acids, which led to interferences of 0.03 mg g−1. Therefore, up to a humic acid content of 10%, there is no bias above the LOQ. However, in humic acid–rich samples such as peat, an overestimation of PET cannot be excluded. Apart from that, however, the method is suitable for quantification of PET in complex environmental samples such as sediment, sewage sludge, or compost.

Table 2 Interferences by organic matrixContamination during sample preparation

To detect a contamination caused by atmospheric deposition [16], six opened reaction vessels were displayed for 14 days at different sites (fume hood and lab bench) in the laboratory where the sample preparation was conducted. PET levels ranged from < LOD up to 996 ng (absolute mass) or 20 µg g−1 (based on a sample weight of 50 mg) (ref. Table S2). However, this was a worst-case scenario: If short exposure times were applied (few minutes compared to 14 days), the risk of contamination was found to be negligible. The background level measuring heated sea sand was always < LOQ (ref. Table S3).

Recoveries and reproducibility

To check whether different matrices have effects on the depolymerization efficiency and reproducibility, spiking experiments were performed in an SPM sample at two different concentration levels (0.01 and 0.5 mg g−1). For both spike levels, recoveries of 96% were achieved. To check the reproducibility of the analytical method, the non-spiked and spiked samples were analyzed in eight replicates over the whole analytical method (shown in Table 3 and Table S4). Relative uncertainties ranged from 8 to 16%, highlighting a high reproducibility and reliability of the analytical method.

For the lower spike level, the statistical uncertainty was relatively high (36%), which was probably caused by the spiking of the authentic sediment sample, which already contained a quantity of PET in the range of the lower spike level. The statistical uncertainty was elevated in the non-spiked sample, probably caused by an inhomogeneous distribution of the PET contamination. Due to the propagation of the uncertainties of the results, the uncertainty of the recovery is elevated.

Table 3 Reproducibility and recoveries for PET in SPM at two spiking levelsAnalysis of different PET products

To evaluate whether methanolysis efficiency is affected by PET type and size, several PET products were tested as well as one sample of poly(butylene adipate terephthalate) (PBAT) (Table 4). Recoveries for all PET types ranged between 87 and 117%. Even the coarse PET particle (> 300 µm) was quantitatively converted to DMT, confirming that the analytical method is not limited to the smaller size fractions. However, particles of this size could cause inhomogeneity issues and samples should be homogenized as far as possible prior to analysis.

For PBAT, a “PET” recovery of 28 ± 16% was determined (i.e., 1 µg of PBAT was transformed to an amount of DMT that would equate to 0.28 ± 0.16 µg of PET). The partition of terephthalate in the PBAT sample was determined by NMR spectroscopy (see SI: Characterization of PBAT) to be 22 ± 2% by mass, resulting in a conversion factor for PBAT into PET of 0.29 ± 0.03. Therefore, the recovery of the actual terephthalate content was 97 ± 21%.

This experiment revealed that the developed method cannot distinguish between PET, PBAT, and other terephthalates. The determined PET concentration should therefore be understood as the sum of concentrations of all terephthalate-containing polymers (“cluster” [17]). A non-polymeric example of a terephthalate that might interfere with PET quantification is the plasticizer di(2-ethylhexyl) terephthalate [18], which is increasingly produced as a phthalate-free plasticizer alternative. However, its production volume is significantly lower than that of PET, and thus a lower environmental concentration can be expected.

Table 4 Recoveries for different PET productsAnalysis of environmental and water samples

Sediments, suspended particulate matter, compost, sewage sludge, and house dust as well as mineral water bottled in single-use PET container and tap water were analyzed with regard to their contamination with PET MP. The results are summarized in Table 5, with a detailed list in SI Tables S5–S11.

Table 5 PET concentrations in environmental samples

PET concentrations in SPM and sediments lie in the same range as reported for PVC (8–220 µg g−1) and PP (32–55 µg g−1) in previous studies, but lower than PE (30–1000 µg g−1) [1, 6]. Compost and sewage sludge showed slightly higher concentrations in the same range as reported in literature [11,12,13].

Highest PET concentrations (up to 5.7% by mass) were found in indoor dust. High amounts of MP—especially fibers—were reported in previous studies [12, 19,20,21]. PET is a common synthetic fiber in textiles such as clothes and carpets. Thus, occurrence of PET in indoor dust is not surprising. However, these high amounts in indoor dust demonstrate the risk of possible sample contamination and the necessity of measures to prevent them.

Due to the enrichment of 9 L and the high sensitivity of the method, PET could be quantified in bottled waters, too. The detected PET concentrations were from equal to up to ten times as high as in tap water and ultrapure water (mean 32 ± 15 ng/L), which were used as blank controls. Contamination of bottled water by microplastics was reported in previous studies [22,23,24,25]. However, measured concentrations were much lower than reported in literature (0.2 to 6 µg L−1 [26] and 100 to 3000 µg L−1 [27]).