Unmixing of mixed oil using chemometrics. (February 2016)

• Record Type: Journal Article
• Authors: Zhan, Zhao-Wen; Zou, Yan-Rong; Shi, Jian-Ting; Sun, Jia-Nan; Peng, Ping'an
• Is Part Of: Organic geochemistry. Volume 92(2016:Mar.)
• Page Start: 1
• Page End: 15
• Publication Date: 2016-02
• Subjects: Mixed oil -- Unmixing -- Chemometrics -- Alternating least square; Organic geochemistry -- Periodicals; Biogeochemistry -- Periodicals; Géochimie organique -- Périodiques; 553.205
• DOI: 10.1016/j.orggeochem.2015.11.006
• Languages: English
• ISSNs: 0146-6380

Abstract

Highlights: Changes of concentrations and biomarker ratios in mixed oils investigated. Biomarker concentrations are more suitable for chemometrics to de-convolute mixed oils than biomarker ratios. The number, fractional contributions and compound compositions of end-members in mixtures can be calculated by ALS-C. Abstract: In this study, chemometrics was used to unmix a set of oil samples that had been mixed in the laboratory using three end-member oils. It was shown that the concentrations of individual compounds in the mixed oil varied linearly with the fractional contribution of each end-member oil. However, biomarker ratios in the mixed oils varied non-linearly with the amount of each end-member oil. This study demonstrates that concentrations and ratios of biomarkers yield different results when de-convoluting mixed oils. Concentrations of biomarkers are therefore more suitable than the biomarker ratios for unmixing mixed oils. Alternating least squares of biomarker concentrations (ALS-C) provides an excellent way to calculate the number, proportions, and compound compositions of the end-members in mixed oil samples. The ALS-C results are accurate, regardless of whether end-member oils are included in the sample set. The biomarker ratios of end-member oils cannot be directly obtained by ALS, but can instead be calculated using related compound concentrations computed by ALS-C. This method should be applied and verified widely using actual geochemical data.