Assessing sample extraction efficiencies for the analysis of complex unresolved mixtures of organic pollutants: A comprehensive non-target approach

Abstract

The comprehensive extraction recovery assessment of organic analytes from complex samples such as oil field produced water (PW) is a challenging task. A targeted approach is usually used for recovery and determination of compounds in these types of analysis. Here we suggest a more comprehensive and less biased approach for the extraction recovery assessment of complex samples. This method combines conventional targeted analysis with a non-targeted approach to evaluate the extraction recovery of complex mixtures. Three generic extraction methods: liquid-liquid extraction (Lq), and solid phase extraction using HLB cartridges (HLB), and the combination of ENV+ and C8 (ENV) cartridges, were selected for evaluation. PW was divided into three parts: non-spiked, spiked level 1, and spiked level 2 for analysis. The spiked samples were used for targeted evaluation of extraction recoveries of 65 added target analytes comprising alkanes, phenols, and polycyclic aromatic hydrocarbons, producing absolute recoveries. The non-spiked samples were used for the non-targeted approach, which used a combination of the F-ratio method and apex detection algorithm. Targeted analysis showed that the use of ENV cartridges and the Lq method performed better than use of HLB cartridges, producing absolute recoveries of 53.1 ± 15.2 for ENV and 46.8 ± 13.2 for Lq versus 19.7 ± 6.7 for HLB. These two methods appeared to produce statistically similar results for recoveries of analytes, whereas they were both different from the produced recoveries via the HLB method. The non-targeted approach captured unique features that were specific to each extraction method. This approach generated 26 unique features (mass spectral ions), which were significantly different between samples and were relevant in differentiating each extract from each method. Using a combination of these targeted and non-targeted methods we evaluated the extraction recoveries of the three extraction methods for analysis of PW.

The effect of extraction methodology on the recovery and distribution of naphthenic acids of oilfield produced water

“…2,13 For example, in our previous study we demonstrated that the choice of the extraction procedure changes the explored chemical space of the sample. 18 In that study even though two out of three extraction methods showed similar performance for the surrogate chemicals, more detailed chemical characterization revealed substantial differences among tested extraction methods. However, that study was focused on the volatile and semi-volatile fraction of PW.…”

“…The details of the extraction procedure for all three methods are provided elsewhere. 18 In short, the Lq method was the dichloromethane (DCM) extract of the acidified PW, repeated three times, with a final volume of 2 mL. A solution of 1N hydrochloric acid was used for acidification of the PW samples.…”

“…This combination of eluents was previously shown to be effective for extraction of analytes with a wide range of chemical and physical properties in complex samples. 18 Three procedural blanks were generated for each extraction method. For Lq method, these blanks were the extract of the glassware using a mixture of DCM and a 1N solution of HCl.…”

Machine learning combined with non-targeted LC-HRMS analysis for a risk warning system of chemical hazards in drinking water: A proof of concept

“…producing false positives). Similarly to the trend analysis, given the dependency of the explored chemical space on the analysis conditions [14,32], the training sets are specic to a sample set and analysis conditions. Therefore, a good understanding of both the matrix and the analytical instrument is crucial to the success of this approach.…”

“…To deal with such large and complex datasets the analysts have to isolate the environmentally relevant features from the generated features lists (i.e. prioritization) [7,9,13,14]. Prioritization may be performed using the intensity of the features and/or based on the statistical signicance of those features when compared to the samples from dierent origins, for example [7,13].…”