Exploring two-trace two-dimensional (2T2D) correlation spectroscopy as an effective approach to improve accuracy of discriminant analysis by highlighting asynchronous features in two separate spectra of a sample.

This study aims to demonstrate two-trace two-dimensional (2T2D) correlation spectroscopy as an effective tool for improving the accuracy of discriminant analysis. Because 2T2D correlation analysis allows sensitive capturing of asynchronous spectral behaviors between two compared spectra of a sample, the subsequent asynchronous correlation features are expected to reveal more sample-to-sample characteristics and discriminants than the original spectral feature. Initially, near-infrared (NIR) spectroscopic authentication of pure olive oil was performed using the spectra collected at 20 °C and 41 °C. When the 2T2D slice spectra of the samples were used for the discriminant analysis, the authentication accuracy reached to 100%, while became degraded in the cases of using the spectra collected either at 20 °C or 41 °C. Furthermore, a simple strategy of utilizing the average spectrum of one sample group as the reference spectrum in the 2T2D correlation analysis was proposed for two-group discrimination and evaluated for the NIR identification of the geographical origins of agricultural products (milk vetch root (MVR) and perilla seed samples). Because the average spectrum of one sample group was used for comparison, dissimilar constituent compositions of the samples in another group were better observed, thereby improving the accuracy of discrimination of the geographical origins of the samples in both cases. The overall results demonstrated that 2T2D correlation analysis is effective for highlighting the minute asynchronous spectral features of a sample and can be expanded for diverse vibrational spectroscopy-based discriminant analyses.