Near Infrared Spectroscopy and Chemometrics Studies of Temperature-Dependent Spectral Variations of Water: Relationship between Spectral Changes and Hydrogen Bonds

Abstract

The present study aims to provide new insights into the temperature-dependent spectral variations in the near infrared (NIR) region of the spectrum of water by comparing chemometrics with spectroscopic analysis. Fourier transform (FT)-NIR spectra of water in the 9000–5500 cm⁻¹ region have been measured over a temperature range of 5–85°C. The observed spectral changes have been analysed by both chemometrics, such as multilinear regression (MLR), principal component regression (PCR) and partial least squares (PLS) regression, and spectroscopic data analyses, such as second derivative, difference spectra and curve fitting. The second derivative of the NIR spectra of water suggests that an intense feature around 6900 cm⁻¹, due to the combination of antisymmetric and symmetric stretching modes of water, consists of at least five component spectra. Each component spectrum may be ascribed to the water species with no, one, two, three and four hydrogen bonds. Curve fitting has been performed for the 6900 cm⁻¹ band and it has been found that the species with no hydrogen bonds increase largely with temperature, while those with more than two hydrogen bonds decrease. The temperature of water has been predicted by use of MLR, PCR and PLS regression. PCR and PLS regression loadings plots for Factor 1 of the models for the prediction of the temperature of water are almost identical with the difference spectrum of water between 5 and 85°C; both the loadings plots and the difference spectrum reflect strongly the changes in the hydrogen bonds of water. Loadings plots of Factor 1 of the PCR and PLS regression models are very similar to each other. It is very likely that since the temperature-dependent spectral variations of water in the NIR region are very regular, and the spectra have only very small noise and baseline changes, PCR and PLS regression select nearly identical factors.