Integrated reflection-FTIR and multivariate partial least squares approach for rapid and accurate assessment of total organic carbon concentration in shale

Abstract

Spectral data of 193 shale samples collected from Cleveland and Wessex Basins in the United Kingdom were generated from the mid-infrared (MIR) region of the electromagnetic spectrum using two types of Fourier transform infrared (FTIR) spectroscopic techniques, attenuated total reflectance (ATR) and diffuse reflectance infrared fourier transform (DRIFT). Spectral data acquired using these techniques were combined with partial least square regression (PLSR) to calibrate models capable of rapid, routine, and accurate quantification of total organic carbon (TOC) concentration in sediment. At the validation stage, DRIFT presented high correlation coefficient (R2 = 0.96) between LECO measured and model-predicted TOC concentration relative to ATR (R2 = 0.89). Further optimisation of the DRIFT model with a balanced data set offered an improved correlation coefficient (R2 = 0.99) at the calibration and validation stages. Prediction of TOC content of shale samples with unknown values using the optimised model presented good correlation and low modelling error (R2 = 0.93; RMSE = 1.5 wt %): an indication that MIR-DRIFT-PLSR has the potential for accurate and precise quantification of sediment TOC concentration. The presence of organic functional groups and mineral absorption bands observed on the MIR spectra renders the technique indispensable for real-time simultaneous characterisation of TOC and minerals in Gas Shales and other formations during exploration.