Multi-sensor approach for modeling rare earth elements within the lignite waste dumps

Due to technological advances demand for rare earth elements (REEs) have been increased. However, exploration or secondary recovery of these elements has not been done properly except for a few countries in the world that supply REEs, such as China. Many studies using inductively coupled plasma mass spectrometry (ICP-MS) have been conducted on analyzing, and finding relationships of the REEs with other elements. Although these laboratory measurements are accurate, they are expensive and time-consuming. Also, a methodology or a model that helps geologists estimate the REE contents in situ is missing. Therefore, we combine the results of ICP with Fourier-Transformed Infrared (FTIR) spectroscopy to model and predict the REE contents within soil samples of mine waste dumps. The FTIR is a portable instrument that has shown many potentials in measuring mineral and organic material absorption features within the Mid-IR range.

Several studies have indicated the potential of lignite mines in the enrichment of REE minerals. Therefore, we have chosen two lignite mines: Profen and Schlenheein within Germany.  During the Miocene, different sub-tropical plants were developed within catchment basins. The decreased plants were covered by soil and glacial during the ice age. Then, the decreased plants sunk into swamps and decomposed into peats. Two parameters of ground heat and pressure led to the H, O, and nitrogen removal and formation of coal seams at the end of the Tertiary.

From the Profen area as our model training area, some of the thirteen samples (70% of the data used for training and 30% for testing), and from the Schlenheein area as our testing area, some of the hundred samples (all data were used for testing and validation) were collected from different waste dumps. The samples were used to collect the FTIR spectra and ICP-MS, and X-ray Diffractograms (XRD). The FTIR spectra were interpreted and validated with the XRD outputs. A matrix consisting of FTIR wavenumbers and their corresponding reflectance value, and ICP elemental content were created. The REEs were considered as the target parameters for modeling and FTIR wavenumbers were used as input parameters for modeling using stepwise multiple linear regression (SMLR). The SMLR models were investigated by checking the corresponding molecular bands assigned to the wavenumbers within the models and ensuring mineralogical sensibility. For the evaluation of the models, coefficient determination (R²) and root-mean-square error (RMSE) were computed. Finally, the models were applied on the testing data of the Profen area to assess their performances. Finally, the models were applied on the Schlenheein area and assessed.

The results showed that Ti has a high correlation with Y and other REEs (approximately 0.7). Also, the presence of titanite was confirmed by the XRD as titanite. Therefore, titanite is the main source of REEs within the waste dumps which was identifiable via the FTIR wavenumbers. The Profen Y model showed an R² of 0.6 for training and testing and 0.55 for the Schlenheein area. The models showed that the FTIR is a powerful tool to quantitatively predict REE contents with an acceptable R² and RMSE.