Near infrared thermometry of carbon dioxide and water using chemometric methods

Near-infrared spectroscopy (NIR) can advantageously probe larger sample volumes than mid-infrared spectroscopy due to the low absorption coefficients of overtones and combination bands. However, NIR spectra are generally complex, necessitating chemometric methods to disentangle spectral features. In this study, we apply principal component analysis (PCA) and partial least squares (PLS) regression to NIR spectroscopy to understand temperature dependence of two widely used solvents, liquid carbon dioxide and water that were selected based on their very different intermolecular interactions. Data were collected in real time in an autoclave across a wide temperature range where little NIR data are available today, i.e. between 1 and 40 °C (CO₂) and between 25 °C and 185 °C (water). This continuous acquisition provided sufficient spectral information to obtain robust temperature models. PCA identified the dominant spectral trends following the temperature variations, while PLS regression allowed to obtain accurate temperature prediction from the spectra. Precise NIR thermometry using PLS based models is reported for both molecules.