Analysis of Near-Infrared Spectral Properties and Quantitative Detection of Rose Oxide in Wine

This study aims to investigate the near-infrared spectral properties of Rose Oxide (4-Methyl-2-(2-methyl-1-propenyl) tetrahydropyran) in wine, establish a quantitative detection, and build relationships between the chemical groups of Rose Oxide and near-infrared characteristic bands, so as to provide ideas and references for the near-infrared detection of a low-content aroma substance in wine. In total, 133 samples with different wine matrices were analyzed using Fourier transform–near-infrared (FT-NIR) spectroscopy. Min–max normalization (MMN), principal component analysis (PCA), and synergy interval partial least squares regression (Si-PLSR) were used for pre-processing, outlier rejection, analysis of spectral properties, and modeling. Finally, the quantitative detection model was established using the PLSR method and the wine sample containing Rose Oxide was verified externally. Eight subintervals (4000–4400 cm⁻¹, 4400–4800 cm⁻¹, 5600–6000 cm⁻¹, 6000–6400 cm⁻¹, 6400–6800 cm⁻¹, 6800–7200 cm⁻¹, 7200–7600 cm⁻¹, 8400–8800 cm⁻¹) were determined as the characteristic band intervals of Rose Oxide in the NIR region. Among them, 5600–6000 cm⁻¹ was assigned to the first overtone C–H stretching in tetrahydropyran ring and methyl as well as the combination C–H stretching of the CH₃ function groups, 6000–6400 cm⁻¹ was assigned to the first overtone C–H stretching of the C–H=group and the combination C=C stretching in isobutyl, and 8400–8800 cm⁻¹ was assigned to the second overtone C–H stretching and C–O stretching in tetrahydropyran ring as well as the C–H stretching vibration in methyl. In addition, 4000–4800 cm⁻¹, 6400–6800 cm⁻¹, and 7200–7600 cm⁻¹ were assigned to the C–H stretching vibration, while 6400–7600 cm⁻¹ was assigned to the C–O stretching vibration. The training result showed that the calibration model ($r_{cv}^2$ of 0.96 and $RMS{E}_{CV}$ of 2.33) and external validation model ($r_{cv}^2$ of 0.84 and $RMS{E}_{CV}$ of 2.72) of Rose Oxide in wine were acceptable, indicating a good predictive ability. The spectral assignment of Rose Oxide provides a new way for the NIR study of other terpenes in wine, and the use of the established Si-PLSR model for the rapid determination of Rose Oxide content in wine is feasible.