Optimization of the extraction and derivatization procedure
Very few procedures for the determination of benzotriazoles by GC in combination with single quadrupole MS have been published23,31,33,35,37. Based on previous literature reports, an acetylation reaction was selected for the derivatization of BTs in this work, which was carried out simultaneously with the extraction process. A careful impact analysis has been carried out for the following process conditions: the type and volume of the extraction solvent, the volume of acetic anhydride (derivatization reagent), the addition of a buffer salt and the time of simultaneous extraction and derivatization. n-Hexadecane, 1-undecanol, chloroform, carbon tetrachloride, toluene and chlorobenzene were tested as potential extraction solvents. The characteristics of selected solvents are included in Table S1 (Supplementary Material). The responses obtained for each analyzed compound when different solvents were used for extraction are shown in Fig. 1A. The volume of each of the solvents used for the extraction was 100 µL. Optimization was performed for milli-Q water to which a mixture of analytes was introduced, with a concentration of 10 µg/L each. Other conditions of the process were as follows: 100 µL of acetic anhydride, five minutes extraction time. The same conditions were used in all optimization experiments unless otherwise stated.
Chlorinated solvents with a density greater than that of water were selected for the tests (see Table S1, Supplementary Material), enabling easy removal of the organic layer after extraction by using test tubes with a conical bottom36. n-Hexadecane and 1-undecanol make it possible to separate from the aqueous solution after extraction by solidifying a floating solvent drop, due to the corresponding melting points, which are 18oC and 13oC, respectively38. Aromatic solvents were selected due to their structural similarity to analytes and the possibility of forming π-π interactions, which may positively affect the extraction efficiency33. As expected, the highest extraction efficiency was achieved with the use of chlorobenzene and toluene, while the area of the chromatographic peaks recorded when the extractant was chlorobenzene was 13 to 22% higher than in the case of toluene. It can also be seen that for chlorobenzene, the lowest standard deviations of the recorded peak areas were recorded. When making the selection, the toxicity of solvents was also taken into account, which was against the chlorobenzene. In contrast, the toxicity of chlorobenzene is lower than that of other organochlorine solvents used in microextraction techniques. The use of toluene as the extractant requires additional operations to separate it from the extracted matrix. Moving to other vessels for this purpose is associated with the risk of losing some of the solvent or recontaminating the sample. At the same time, all additional operations increase the exposure of the analyst and the environment to the contact with the solvent. Taking all the above into account, chlorobenzene was selected as the optimal solvent for the extraction.
Optimal solvent volume was selected on the basis of tests performed at 40, 60 and 80 µL. The results of the conducted experiments are presented in Fig. 1B. As expected, there was clearly an inverse relationship between the volume of the solvent and the area of the obtained analyte peaks. Therefore, it was found that the volume of 40 µL is optimal for the conducted experiments. The solubility of chlorobenzene in water is approximately 500 mg/L, so the theoretical amount of solvent that can be recovered after extracting a 5 mL water sample is 37.5 µL. After extraction, it was possible to collect about 25 µL of chlorobenzene, which made it possible to carry out several replications of the analysis using the autosampler of the GC-MS device. In order to make sure that the determined optimal volume could be used during the wastewater analysis, additional experiments were carried out using influents as well as the wastewater from different treatment stages. For influents, it was found necessary to use a solvent volume of 80 µL to obtain 15-20 µL of extract that could be analyzed by GC-MS. It is related to the presence of a rich matrix in wastewater, including macromolecular compounds (proteins, fats and carbohydrates), i.e. substances with emulsifying properties, increasing the solubility of chlorobenzene39. Therefore, it was decided to use a volume of solvent equal to 80 µL in all subsequent experiments. However, when the developed method will be used for samples of less polluted water, including surface water, groundwater, and purified sewage, it is possible to use a chlorobenzene volume of 40 µL.
The optimization results for acetic anhydride volume, ionic strength and extraction time are shown in Fig. 2. The effect of acetic anhydride volume on the extraction efficiency was investigated in the range from 60 to 250 µL and 125 µL was selected as optimal. It turned out to be sufficient also when the experiments were repeated with samples of raw sewage enriched with analytes at a concentration of 100 µg/L.
Many studies have been carried out and described in the literature on the selection of the appropriate salt or buffer solution for liquid-liquid microextraction36,40. As a result, it was found that the optimal choice in this case are phosphoric (V) acid salts with buffering properties. In microextraction methods, especially in USAEME, the use of sodium bicarbonate is disadvantageous because the formation of CO2 bubbles makes it difficult to separate the extract from the aqueous matrix41. However, the influence of the ionic strength on the process of benzotriazole microextraction is unclear, as the available literature gives completely opposite findings26,33. In this study, the effect of salt addition ranging from 0 to 10% was investigated. The obtained results do not show an unequivocal relationship between the obtained analyte peak areas and salt concentration. Similar results were obtained for concentrations of 0, 2.5 and 5%. In the case of concentrations of 7.5 and 10%, higher peak areas were obtained for BT and 4MBT, and lower for 5MBT. In order to definitively settle this issue, the research was repeated for 0 and 10% concentrations with the use of influent wastewater with addition of benzotriazoles. Studies have shown that in the case of a very complex matrix, which is untreated wastewater, the addition of buffering salt causes precipitation during extraction, which makes it difficult to take the appropriate volume of extract. Therefore, in further works, no salt was added to the extracted matrix. A similar solution was selected as optimal also when using the DLLME technique33.
The relationship between the time of simultaneous extraction and derivatization, and the extraction efficiency was studied in the range of three to 12 minutes. Extending the simultaneous extraction and derivatization time from 3 to 5 minutes results in an increase in surface areas by about 10%. The surface areas obtained after 12 minutes of the process, for all tested compounds, were about 30% lower than those obtained after 5 minutes. The decrease in analyte surface areas with the extension of sonication time indicates that acetyl derivatives of benzotriazoles are likely to partially hydrolyze to free forms upon prolonged contact with the aqueous phase. This phenomenon has already been observed in the case of other polar compounds38.
Method validation
Analytical parameters including linearity, precission, LoD, and limit of quantification (LoQ) as well as method recovery were investigated under established optimal conditions. Calibration curves were obtained by spiking the ultrapure water with seven concentration levels, between 0.05 and 10 µg/L and performing the extraction and GC-MS analysis. In the case of 5ClBT two chromatographic peaks were recorded. This situation is probably related to the formation of two isomeric acetyl derivatives of 5ClBT, substituted at the 1 and 3 positions of the triazole ring42. Calibration curves for this compound were plotted and validation parameters were determined based on peak 1, peak 2 and the sum of the peaks. The validation data are summarized in Table 1. Calibration curves were linear within the studied concentration ranges with coefficients of determination (r2) ≥ 0.9900 for all target compounds.
The LoD was established as the concentration giving a signal-to-noise ratio (S/N ratio) of 3 and it was measured in chromatograms registered for the solutions with lowest analytes concentration. The LoQ was calculated as the concentration corresponding to an S/N ratio of 10. LoDs and LoQs values for the tested benzotriazoles ranged from 0.1 to 1.8 ng/L and from 0.4 to 5.9 ng/L, respectively. The lowest detection and quantification limits were recorded for UV326, UV329 and 5MBT.
The precision of the determinations was established on the basis of CV calculated for developed model (calibration curve). The CV for each point was calculated as the ratio of the root mean squared error to the mean of the concentration calculated on the basis of the model43. The CV given in Table 1 was calculated as the average percent value of the variance coefficients determined for seven points of the calibration curves. For the individual compounds tested, mean CV values ranging from 4.7 to 9.8% were obtained. As expected, the measurements for higher concentrations are characterized by the highest precision. The mean CV value for all tested compounds at a concentration of 10 µg/L was 4.7%, while for the concentration of 0.05 µg/L it was almost 15%. Recoveries for each compound were determined at two concentration levels: 0.2 and 2 µg/L. They were calculated comparing nominal concentration with the value determined based on the calibration curve. The recovery values were between 81 and 116% for the lower concentration, and between 87 1nd 118% for the higher value.
Matrix effect
Municipal and industrial wastewater are water matrices with a high degree of pollution and properties strongly differing from those of pure water. A complicated matrix may affect the surface area of the obtained analytical signals as well as the location and course of the baseline. It is assumed that the influence of the matrix at the level of -20 to 20% is considered acceptable and determinations can be made on the basis of a calibration performed on a simplified matrix. With stronger disturbances, matrix-matched calibration must be used44. The tests of the resistance of the method showed that the discrepancies of the received signals exceed the determined range. The highest exceedances were recorded when the matrix was influents. Therefore, the developed method was validated with the use of six different real matrices (see Table S2 and Table S3, Supplementary Material). For all municipal wastewater matrix, good linearity was obtained, expressed as r2 value above 0.99. The mean recovery deviation from the value of 100% was the highest for influent wastewater and was approximately 18%. In the case of the remaining matrices, it ranged from 8 to 13%, i.e. it did not differ significantly from the values determined for water. The CV values did not exceed 10%. It was probably influenced by the higher concentrations within the curve. The average sensitivity of determinations for the group of the target compounds, expressed by the LoD value, in the case of municipal wastewater as matrices, ranged from 6 to 14 ng/L. In the case of industrial wastewater as a sample matrix, the deviation of validation parameters from those registered for water is higher than in the case of municipal wastewater. In this case, average values of the difference in recovery from 100% equals 25% for influents and 12% for wastewater after flotation. Higher values of CV, and LoD were also recorded. It can be seen that the deterioration of the recovery and the reduction of sensitivity, is quite well correlated with the contamination of the matrix used, expressed by the BOD, COD, suspension, nitrogen and phosphorus concentrations (see Tables S4, S5, S6 Supplementary material).
Comparison with other analytical procedures
Due to its high polarity and limited volatility, liquid chromatography is the most commonly used in BTs analysis, usually combined with MS or MS/MS detection. The limited availability of this type of equipment as well as the high cost of determinations and large production of waste solvents are significant obstacles of LC-MS/MS technique. GC-MS systems, especially those with one quadrupole, are currently common equipment in laboratories conducting environmental determinations. At the same time, GC-MS analyzes are cost-friendly, easy to perform and environmentally friendly due to the small amount of waste generated. The development of derivatization in the matrix in recent years facilitated the use of GC in the analysis of compounds even with high polarity. Table 2 presents a comparison of the developed method with other solutions used in the determination of BTs. Due to the fact that in the literature from recent years there are comparisons of BTs determination methods, only procedures based on gas chromatography and various isolation techniques were taken into account22,26,29. Comparing the developed procedure with others described in the literature, it can be stated that the validation parameters determined by us are similar or better than in the case of other determinations performed with the GC-MS and GC-MS/MS techniques. The comparison of isolation techniques shows that SPE, DLLME and USAEME allow determinations with similar accuracy, precision and sensitivity, while the use of SPME is associated with the deterioration of the validation parameters of the procedure. It should be emphasized, however, that the procedures based on the use of SPE are multi-stage (conditioning of columns, sample application, elution, concentration of the eluate) and require a large volume of samples (even 2.5 L, on average 1L) and the use of large volumes of organic solvents (up to 61 mL per one repetition of the procedure)23. These facts justify the growing popularity of liquid-liquid microextraction techniques. The DLLME-GC-MS and USAEME-GC-MS procedures have similar analytical characteristics (with a slight advantage of the first technique), while the 20 times higher consumption of organic solvents in the DLLME technique (sum of the extraction and dispersion solvent volumes) should be considered as an advantage of USAEME. A significant improvement of the proposed method, compared to the DLLME-based procedure, is the simple, one-step recovery of the solvent after extraction, which is associated with a lower risk of sample contamination and analyte loss33.
Wastewater analysis
The developed USAEME-GC-MS procedure was used for the simultaneous determination of compounds from the LMBT and BUV groups in wastewater from two municipal wastewater treatment plants and in the outflow of MPP. Fig. 3 show chromatograms recorded for the influent and effluent wastewater from WWTP A. Table 3 summarizes the concentrations of the target LMBTs and BUVs in samples of municipal wastewater from WWTP A and WWTP B and in dairy wastewater from MPP. BT, 4MBT and 5MBT determinations were performed on six samples of municipal influents and six samples of effluents (five from WWTP A and one from WWTP B). 5ClBT, UV326, and UV329 determinations were carried out in four samples of each type from WWTP A. Average concentrations of LMBTs in influent wastewater, along with deviations, ranged from 0.055±0.007 µg/L for 5ClBT (N=2, two values below the LOD) to 2.205±2.335 µg/L for BT. Among the methyl derivatives of BT, the higher concentration of 1.117±2.058 µg/L was recorded for 4MBT and for 5MBT it was 0.158±0.136 µg/L. For BUVs, the mean concentrations in municipal influents were 0.190±0.082 µg/L and 0.167±0.038 µg/L for UV326 and UV329, respectively (N = 3, one value below LOD). In the treated wastewater, the average concentrations of BT (N=6), 4MBT (N=6), 5MBT (N=4) and 5ClBT (N=2) were 1.065±0.907 µg/L, 0.150±0.171 µg/L, 0.178±0.215 µg/L, and 0.010 µg/L, respectively. The mean concentrations of UV326 (N=3) and UV329 (N=2) were 0.040±0.036 µg/L and 0.050±0.028 µg/L, respectively. The LMBTs content in municipal influents determined in this study is similar to the values obtained during the research conducted in Spain by Dominguez et al.31. Other results that can be found in the literature indicate a higher content of these compounds, reaching 13 µg/L for BT and over 7 µg/L for other LMBTs22,45,46. The content of LMBTs in municipal effluents similar to that recorded by us was obtained by Liu et al. and Casado et al. for samples from Australia and Spain22,33,34. The remaining literature reports give higher concentrations of these compounds in effluents, reaching 10 µg/L45. The content of BUVs in municipal influents reported in the literature is over ten times higher than the values recorded by us34.
Fig. 4 shows the range and mean ML of target compounds flowing with influents into WWTP A and introduced with effluents from WWTP A into the aquatic environment. Fig. 5 shows the range and average values of the removal efficiency (RE) of individual compounds in the WWTP A. When no compound was detected in a specific matrix, the appropriate LOD value was inserted into the calculation of ML and RE. The ML values for each target compound was estimated according to Eq. (1). The average ML values calculated for influents ranged from approximately 8 mg/day/1000 inhabitants for 5ClBT to almost 600 mg/day/1000 inhabitants for BT. For the remaining compounds, these values were at the level of several dozen mg/day/1000. The lowest mass load introduced into the environment was recorded for 5ClBT and it was 2.2 mg/day/1000 inhabitants. 269 mg/day/1000 inhabitants was the ML for BT, which was the highest recorded value. The average ML value for the remaining compounds ranged from a few to several mg/day/1000 inhabitants.
Efficiency (RE, %) of the removal of LMBTs and BUVs during the treatment process depends on its tendency to be absorbed by activated sludge as well as its biodegradability and physiochemical properties (volatility, polarity, water solubility). RE is defined as negative or low when RE < 20%, moderate when the RE ranged between 20.1% and 70% and high when the RE exceeded 70.1% 47. As shown in Fig. 5, the average RE value obtained for target compounds allows the removal efficiency of BT, 4MBT and 5MBT to be classified as moderate and 5ClBT, UV326, and UV329 as high. In the case of BT, 4MBT, 5MBT, in some sampling campaigns higher concentrations were observed in effluents than in influents, i.e. RE values below zero were obtained. “Negative removal” has been previously observed and it could be explained by the varying sorption of the target compounds in activated sludge and their subsequent desorption by successive portions of wastewater48. The reason may also be the fact that the influents and effluents were collected at the same time, while the treatment process in WWTP A takes about 48 hours, so the effluents come from a completely different portion of influents than taken on a given day49. On the basis of the obtained results it can be concluded that the removal of BTs by activated sludge processes is not sufficient and it is advisable to use additional technologies as the tertiary wastewater treatment.