Application of Multivariate Data Analysis Methods for Rapid Detection and Quantification of Adulterants in Lavender Essential Oil Using Infrared Spectroscopy

Abstract

Lavender, widely cultivated in the Mediterranean region, produces essential oil known for its significant biological activities and is a key component of the perfume industry due to its high levels of Linalool and Linalyl acetate, along with low Camphor content, which contributes to its high cost. However, the market is plagued by adulterated lavender oil, often mixed with cheaper alternatives such as eucalyptus and rosemary. Current detection methods, primarily gas chromatography, are expensive, time-consuming and often fail to detect low levels of adulteration. To address these limitations, this study examines the use of mid-infrared spectroscopy for the detection and prediction of adulteration levels. A set of 105 samples, comprising pure lavender oil and adulterated lavender oil, was prepared in the laboratory. Principal component analysis (PCA), hierarchical clustering ascending (HCA) and K-means clustering were applied to the FT-MIR results for qualitative analysis to effectively discriminate between authentic and adulterated essential oils. For quantitative analysis, partial least squares regression (PLSR) was used to develop accurate calibration models for predicting the percentage of adulteration. The results from PCA, HCA and K-means demonstrated the efficacy of these techniques in detecting adulteration, even at low levels (2%). Calibration models were developed using the PLSR method with different spectral preprocessing techniques to predict the percentage of adulteration, with results indicating that models generated on the raw data and those using MSC (multiplicative signal correction) pre-processing are optimal. In addition, the use of interval-partial least squares (IPLS) variable selection techniques (Forward, Backward) improved the predictive accuracy of the models developed by reducing the number of wavelengths used.