Characterisation of complex perfume and essential oil blends using multivariate curve resolution-alternating least squares algorithms on average mass spectrum from GC-MS.

Quality control of essential oil blends and the discovery of potential adulterations and product fraud is a significant challenge within the natural oil and perfume industry. In this research, total chromatogram average mass spectra (TCAMS), created from the GC-MS three-way raw data, were employed for the characterisation of complex samples of perfumes and essential oil blends. A multivariate approach for curve resolution was used to resolve the TCAMS of pure essential oils within such perfume and essential oil blends. Resolved TCAMS, in combination with unsupervised pattern recognition approaches revealed the distillation grade and origin of used ylang-ylang oils in perfume mixtures. TCAMS resolved from the essential oil blends were used with a supervised machine learning classification model to identify oils, used in creating the blends. Quantification was performed using a multivariate curve resolution approach, resulting in relative errors of prediction lower than 17.84% with root mean square errors of prediction smaller than 3.43.