Prediction of Na- and Ca-montmorillonite contents and swelling properties of clay mixtures using Vis-NIR spectroscopy

The aim of this study was to evaluate whether Na- and Ca-montmorillonite and the swell-indicating properties (i.e., free swell index, water uptake capacity, and cation exchange capacity (CEC)) of clay mineral mixtures can be estimated using visible-near-infrared (Vis-NIR) spectral features. The data regarding four types of reference clay minerals (KGa-1b, kaolinite; IYd, illite; SWy-3, Na-montmorillonite; STx-1b, Ca-montmorillonite) and binary and ternary mixtures of the reference clay minerals (SWy-3/KGa-1b/IYd, and STx-1b/KGa-1b/IYd) with specific mass percentage ratios were used as a calibration dataset. The absorption spectral features could be correlated well with changes in the clay mineral content of the mixtures. The leave-one-out cross-validation results showed that the partial least square regression (PLSR) calibration models produced the best prediction in the following order: Na or Ca-montmorillonite, kaolinite, and illite amounts in the mixtures. The calibration models produced better predictions in the ascending order of CEC, free swell index, and water uptake capacity, regardless of the multivariate statistical methods and type of exchangeable cations in montmorillonite. Validation using an independent dataset indicated that the PLSR model predicted the Na- and Ca-montmorillonite content in clay mineral mixtures and bentonite samples. The results of this study provide the possibility of using Vis-NIR absorption spectral features as a screening tool for predicting the montmorillonite content with different interlayer cations in relatively pure clay soils (e.g., bentonite), if the calibration model is developed using the specific montmorillonite contained in the clay soil of interest.