Smartphone-based handheld Raman spectrometer and machine learning for essential oil quality evaluation.

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Essential oils (EOs) are widely used in various industries.
• Adulteration of EOs with cheaper substances is a significant concern, affecting quality and safety.
• Accurate and rapid detection methods for EO adulteration are crucial.

Purpose of the Study:

• To develop and validate a smartphone-based Raman spectroscopy method combined with machine learning for essential oil analysis.
• To discriminate pure EOs from adulterated ones and identify adulterant types.
• To quantify the levels of specific adulterants in essential oils.

Main Methods:

• Utilized a smartphone-based miniaturized Raman spectrometer for spectral data acquisition.
• Employed machine learning algorithms, including Random Forest and Partial Least Squares Discriminant Analysis (PLS-DA), for classification.
• Applied Partial Least Squares Regression (PLS) for quantitative analysis of adulterants.

Main Results:

• Achieved excellent performance in discriminating pure EOs from adulterated samples using Random Forest and PLS-DA.
• Successfully identified the type of adulteration, including solvents (benzyl alcohol), vegetable oils, and lower-priced EOs.
• Quantified adulterants with a relative error of prediction (REP) between 2.41% and 7.59% using PLS analysis.

Conclusions:

• The developed smartphone-based Raman spectroscopy and machine learning approach offers a fast, reliable, and portable method for essential oil quality control.
• This technology can effectively detect and quantify adulteration in essential oils, safeguarding product integrity.
• The method holds promise for on-site authentication and quality assessment of essential oils.