Chemometrics of meta-clay FT-IR spectra: From heating temperature estimation to meta-clay identification

In this research, chemometric tools (Partial Least-Squares Regression, Principal Component Regression, Linear Discriminant Analysis for example) allowed estimating firing temperatures and to identify the type of meta-clay based on their infrared spectra. The 3 min phases observed in clay thermal degradation were: (i) a dehydroxylation reaction between 400 and 500–600 °C leading to the formation of a non-crystalline meta-clay, (ii) a continuous thermal degradation reaction of the meta-clay involving the Si O bonds between 500 and 850 °C, (iii) the formation of secondary silicates above 850 °C. The 10 Latent Variables of the Partial Least-Squares Regression model led to accurate predictions (Root Mean Squared Error of Prediction = 66.6) and provided a good correlation between predicted and actual temperatures, without any estimation bias related to the type of clay, the presence of other minerals or iron concentration. These models give greater importance to the infrared absorption bands of the bonds involved in thermal degradation: Si-O-Si (1026 and 461 c m − 1 ), Si-O-Al (526 c m − 1 ), Si O or Al O (1095 c m − 1 ), and infrared bands resulting from the formation of meta-clays (486 et 562 c m − 1 ). Finally, a Linear Discriminant Analysis model is proposed in order to identify the type of meta-clay between different types of meta-kaolinites, meta-nacrite and meta-smectite in 96 % of cases. The meta-clays produced by the firing of halloysite, dickite and kaolinite were too close in terms of infrared spectra to be distinguished. • Identification of 3 phases in clay thermal degradation. • PLSR and PCR models led to accurate predictions of cooking temperatures ((RMSEP = 66.6 for PLSR). • Most important infrared bands were Si-O-Si, Si-O-Al, Si-O or Al-O, and infrared bands resulting from the formation of meta-clays. • a LDA model allowed identifying the type of meta-clay in 96 % of cases.