Determining the Concentration of Polycyclic Aromatic Hydrocarbons in Water Using Surface Enhanced Raman Spectroscopy and Kernel Principal Components Analysis Combined with Support Vector Regression

C. Jian,J. Boyan,Zh. Ying,W. Zhenyu

DOI: https://doi.org/10.1007/s10812-021-01161-z

2021-03-01

Journal of Applied Spectroscopy

Abstract:Determining the concentration of polycyclic aromatic hydrocarbons (PAHs) in water is vital for reducing negative effects on human health, such as cancer and malformation. This study proposed an alternative analytical method based on surface enhanced Raman spectroscopy and kernel principal components analysis combined with support vector regression (SVR) for the determination of PAH concentration in water. For this, a dataset containing 300 Raman spectra of polycyclic aromatic hydrocarbon mixtures was made using naphthalene (NAP), pyrene (PYR), and phenanthrene (PHE) with concentrations ranging from 0 to 1000 ppb. In order to improve the effect of the model detection, different pre-processed methods were applied: normalization, multiplicative scatter correction, detrending, standart normal variate transformation, and Savitzky–Golay smoothing. For comparison, partial least squares (PLS) and SVR with the polynomial-kernel were also used. The pre-processing method with the best prediction effect was SNV for all the three substances. For NAP, the optimal correlation coefficient of cross-validation (Rcv), correlation coefficient of prediction (Rpred), RMSECV, RMSEP, and RPD are 0.90, 0.937, 138.9, 117.4, and 2.9 ppb, respectively, while for PYR the optimal Rcv, Rpred, RMSECV, RMSEP, and RPD are respectively 0.881, 0.897, 152.3, 142.8, and 2.3 ppb. For PHE, the optimal Rcv, Rpred, RMSECV, RMSEP, and RPD are 0.980, 0.982, 64.5, 62.9, and 5.3 ppb, respectively. This study provides a new method with a better prediction effect for quantitative analysis of low concentrations of polycyclic aromatic hydrocarbons in water by using surface enhanced Raman spectroscopy.

spectroscopy

What problem does this paper attempt to address?

This paper aims to solve the problem of determining the concentration of polycyclic aromatic hydrocarbons (PAHs) in water. PAHs are a class of aromatic compounds with two or more benzene rings and are widely present in the environment as persistent organic pollutants. Most of these compounds are generated during industrial processes and are directly discharged into the air, while a small amount enters the soil and water bodies. Due to their high molecular weight, PAHs have very low volatility and water - solubility, so they easily enter water bodies by adsorbing on particulate matter and depositing at the bottom of the water. PAHs are difficult to degrade under natural conditions, so their content in water is usually lower than that in sediments, which makes measurement difficult. In addition, PAHs may eventually accumulate in the human body through bio - accumulation in the food chain. However, PAHs are highly carcinogenic and teratogenic, so detecting their concentration is crucial for reducing negative impacts on human health (such as cancer and deformities).

For this purpose, the study proposes a new analytical method based on surface - enhanced Raman spectroscopy (SERS) combined with kernel principal component analysis and support vector regression (SVR) for determining the PAHs concentration in water. This method can not only improve the detection speed and accuracy, but also enable the quantitative analysis of low - concentration PAHs. By using different pre - treatment methods (such as normalization, multiple scattering correction, detrending, standard normal variate transformation, and Savitzky - Golay smoothing) to improve the model detection effect and comparing with partial least squares (PLS) and support vector regression (SVR), the results show that the standard normal variate transformation (SNV) has the best prediction effect for three substances (naphthalene, pyrene, and phenanthrene). Finally, this study provides a new quantitative analysis method with better prediction effect, which is suitable for determining low - concentration polycyclic aromatic hydrocarbons in water using surface - enhanced Raman spectroscopy.