Excitation emission matrix fluorescence spectroscopy and parallel factor framework-clustering analysis for oil pollutants identification

https://doi.org/10.1016/j.saa.2021.119586Get rights and content

Highlights

  • A new approach was proposed for oil identification.

  • EEM spectroscopy and PFFCA were applied.

  • Nontrilinear structures in EEM fluorescence spectra were resolved.

  • Compared PFFCA-LDA, PARAFAC-LDA and PLS-DA methods by statistical parameter.

  • PFFCA-LDA has proven to be the most effective tool.

Abstract

Serious ecological damage can be caused due to increased oil pollution. Identifying the source of oil can inform effective mitigation strategies and policies. A novel method for oil pollutants identification has been presented based on excitation emission matrix (EEM) fluorescence spectroscopy and parallel factor framework-clustering analysis (PFFCA). First, the EEM spectroscopy of the oil samples was measured by a FS920 steady-state fluorescence spectrometer. Second, EEM was analyzed and characterized by PFFCA. A total 90 EEM were decomposed into six components using excitation wavelengths from 260 to 400 nm and emission wavelengths from 280 to 450 nm. Finally, oil samples were classified and matched by using concentration vectors. The results were compared with those obtained by using linear discriminant analysis (LDA) employing parallel factor analysis (PARAFAC) scores, and partial least squares discriminant analysis (PLS-DA). The best classification result was obtained by using LDA employing concentration vectors with 96.7% accuracy. The results indicate that PFFCA-LDA offers a robust approach for the oil samples, which is of great significance in discrimination of oil pollutants.

Introduction

Accidental and deliberate oil spills occur frequently during exploitation, transportation, storage and use [1]. Some negative and long-term impacts on the ecosystem and human health are caused due to the high content of toxic and mutagenic compounds in oil [2], [3]. Accordingly, rapid and reliable methods for oil identification is a constant need in order to determine the source of spills.

Fluorescence spectroscopy has been widely used for forensic analysis of oil products due to its advantages of selectivity and high rapidity and sensitivity [4], [5], [6]. The EEM fluorescence spectroscopy coupled with PARAFAC analysis has been used as a powerful tool to characterize and match oil pollutants [7], [8], [9], [10]. The complex EEM fluorescence spectra of the oil products can mathematically be separated into several pure factors by using PARAFAC analysis [11]. Those pure factors usually indicate the underlying chemical constituents of oil products and can be used to characterize oil products [12], [13].

The fluorescence spectroscopy of oil products mainly corresponds to their polycyclic aromatic hydrocarbon (PAH) profiles [10], [14], [15], [16]. Because the fluorescence characteristics of isomers with similar chemical structure are almost identical [7], it is unlikely that PARAFAC analysis is able to distinguish all the individual compounds present in oil products containing hundreds of PAHs and their homologues. Furthermore, the analysis is further complicated because of the highly variant structure, conformation and heterogeneity, as well as intra- and intermolecular interactions of PAH [17].

Recently, an interesting approach for EEM fluorescence spectroscopy analysis is parallel factor framework-clustering analysis (PFFCA), which was proposed by Chen et al. [18] for the defects of PARAFAC analysis. PFFCA accurately determines the actual component number by eliminating the PARAFAC limitation of requiring trilinear results. Consequently, the nontrilinear structures in EEM fluorescence spectra of environmental samples can be robustly treated by PFFCA.

In this paper, the PFFCA is used for analyzing and characterizing the EEM fluorescence spectra of six types of oils. The aim of this study is to develop a rapid, reliable and accurate fluorescence spectrum analysis method to discriminate oil contaminants.

Section snippets

Parallel factor framework-clustering analysis (PFFCA)

A more detailed description of PFFCA calculations could refer to reference [18] and a brief description was given as follows:

The PFFCA method principally assumes that all of the components are independent [18]. Firstly, the EEM data set is decomposed many factors using the PARAFAC model. That is, if EEM data set is arranged in a three-way array X of dimensions I × J × K (where I is number of samples, J is number of emission wavelengths and K is number of excitation wavelengths). X is decomposed

Sample preparation

Six types of oil were chosen as the experiment samples: diesel (0#), gasoline (92#), gasoline (95#), jet fuel, mineral oil (industrial grade) and lubricating oil. A solvent with a concentration of 0.1 mol/L was obtained by dissolving sodium dodecyl sulfate (SDS) in seawater (Bohai Sea, China). Each oil was weighed by using the precision electronic scale (FA1004, Tianjin Tianma Weighing Apparatus Co., Ltd., China). The actual weight obtained of six oils were listed in Table 1. The primary

EEM fluorescence spectra characteristics of oils

EEM fluorescence spectra of six oils were shown in Fig. 1. It can be observed that Rayleigh scattering appear diagonally in EEM fluorescence spectra of all oils. The Rayleigh scattering does not contain any fluorescence information. Fluorescence of lubricating oil was masked due to the highly intense Rayleigh scattering in Fig. 1f. Consequently, the Rayleigh scattering must be eliminated before processing the EEM fluorescence spectra for analysis. In addition, Raman scattering also exists in

Conclusion

A novel approach for classification and matching of oil samples has been developed and applied in this paper. It is based on three-way decomposition of EEM fluorescence spectra by PFFCA analysis. Six components were extracted from the EEM of oil samples. The component score matrix was used for classification and matching of oil samples. The six types of oil samples were identified with an accuracy of 96.7%. Compared with PARAFAC-LDA (accuracy of 87%) and PLS-DA (not assigned of 33.3%), better

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

Thanks to the financial support provided by the National Natural Science Foundation of China (Grant 61771419, 61501394) and the Natural Science Foundation Project of Hebei Provincial (Grant F2017203220, F2016203155).

References (22)

  • M. Alostaz et al.

    Petroleum contamination characterization and quantification using fluorescence emission-excitation matrices (EEMs) and parallel factor analysis (PARAFAC)

    J. Environ. Eng. Sci.

    (2008)
  • Cited by (0)

    View full text