Geofingerprinting of coltan using handheld spectroscopic devices

Tantalum (Ta) and niobium (Nb) are extracted from many different ore types. The major economic orebody mined for tantalum and niobium consists of tantalite ((FeMn)Ta2O5) and columbite ((FeMnNb2O5)) in a solid-solution mineral commonly referred to as “coltan.” The largest supplying region, The Democratic Republic of the Congo (DRC), is unfortunately also the most ethically compromised source for this critical mineral. Officially, the U.S. government has deemed coltan ore from the DRC a “conflict mineral” [1]. The lack of pragmatic and effective local mining regulation in Central Africa, when coupled with high levels of violence and corruption, has spawned a widespread, illegal, unethical and, most importantly, unsafe economy. This illicit economy is centered around first, illegally mining these valuable conflict minerals and second, smuggling these minerals into the global supply chain. Handheld spectroscopy devices could provide further transparency to these opaque supply chains, giving miners, traders and refiners the ability to authenticate the provenance of the minerals they encounter, and allowing companies to prove compliance with the regulatory agencies attempting to curtail illicit mining. Background Handheld spectroscopy devices allow the operator to collect detailed chemical information from mineral samples. This information includes both bulk compositional data and trace elements. Specifically, laser-induced breakdown spectroscopy (LIBS) and X-ray fluorescence (XRF) handheld devices are of interest in this research, as they are both readily available commercially and capable of collecting data that can later be used for provenance-based classification. Both of these devices have been successfully used to identify minerals through observation of the trace elements. The presence of trace elements, in conjunction with major elements, in a mineral’s spectra can be used to determine a mineral’s geochemical fingerprint [2]. This method can provide both qualitative and quantitative results while also allowing a user to record a plethora of information in seconds [3]. Methods Various coltan samples were acquired, and spectra (LIBS and XRF) were collected from these samples. The collected spectra were then fed into both unsupervised machine learning (UML) and supervised machine learning (SML) classifiers. The UML classifications achieved by these methods were visualized using constellation and dendrogram plots. Further investigations were conducted on the SML classification methods, varying both the number of synthetic spectra and the use of principal component analysis (PCA) to preprocess the spectra. To standardize the spectra collection and to simulate the real-world use of these handheld devices, each type of spectra was collected for one minute. Within the one-minute period, two sets of 10 LIBS spectra could be collected, and two sets of XRF spectra were collected. A handheld LIBS device (SciAps Model Z-300) and a handheld XRF device (Olympus Model Vanta-m) were used in this research. Fig. 1 Hierarchical clustering dendrogram plot, generated using raw XRF sample spectra. The raw XRF and LIBS spectra from samples with known origins were used to train a machine learning multi-class classification algorithm. This algorithm consisted of multiple individual models (random forest classifier, linear regression, support vector classifier, multilayer perceptron). Each individual model attempted to classify the spectra by provenance. The accuracy of each model was recorded, allowing for individual review of how well each SML model was able to classify, and to investigate the ability of each type of spectra (XRF and LIBS) to be individually used to classify the provenance of coltan samples. Finally, synthetic data (a weighted combination of randomly selected spectra from each sample for both LIBS and XRF) were generated. Tests were run using: Only raw XRF and LIBS spectra. Preprocessed (via PCA) XRF and LIBS spectra. A combination of PCA preprocessed XRF and LIBS spectra. Results Unsupervised machine learning. The samples were classified by comparing each sample’s spectra at each individual energy level, recognizing the similarities and dissimilarities. The spectra were then normalized to a minimum intensity of zero and a maximum intensity of one. Hierarchical clustering was then used to generate dendrograms and constellation plots to visualize the clustering/grouping achieved using UML. Figure 1 shows a dendrogram generated using LIBS spectra. Upon review of the dendrogram and constellation plots, some classification of coltan sample provenance can be achieved using UML. This classification, while encouraging, failed to distinctly identify differences between samples with similar provenances, as many South Dakota-based samples were difficult to distinguish from one another. Supervised machine learning. Two different sets of data (either raw spectra or preprocessed using PCA) were used in this research. For the second data type, PCA preprocessing was applied to both the XRF and LIBS spectra. The classifier’s ability to accurately classify the samples, using either of these two data types was recorded, shown in Table 1. Conclusions The efficacy of using both XRF and LIBS spectroscopy methods for provenance determination of coltan single crystals was quantified. The accuracies recorded for each data type when classified using SML confirm the previous research related to geofingerprinting coltan samples. While LIBS alone was unable to achieve 100 percent classification, when used in conjunction with XRF, 100 percent classification was achieved. Both raw XRF and LIBS spectra, alone, provide ample spectral information to classify coltan single crystals to accuracies above 80 percent. When preprocessed using PCA analysis, these accuracies reach approximately 96 percent and higher. The combination of PCA-preprocessed LIBS and XRF was proven to provide 100 percent classification on the samples collected for this research. Selected references 111th Congress, Dodd-Frank Wall Street Reform and Consumer Protection Act, Pub. L. No. 111–203, 124 Stat. (2010) An act to promote the financial stability of the United States by improving accountability and transparency in the financial system, to end “too big to fail,” to protect the American taxpayer by ending bailouts, to protect consumers from abusive financial services practices, and for other purposes, pp 2213–2218 Hark RR, Harmon RS (2014) Geochemical fingerprinting using LIBS. In: Musazzi S, Perini U (eds) Laser-induced breakdown spectroscopy: Theory and applications, Springer Berlin Heidelberg, Berlin, Heidelberg, pp 309–348 Kessinger S (2022) Geofingerprinting coltan using handheld spectroscopy devices. South Dakota School of Mines and Technology, p 83

Geofingerprinting of coltan using handheld spectroscopic devices

Authored by: Samuel Kessinger, Jon Kellar and Prasoon Diwakar

To view this article you must be a member of SME/UCA.

Get your SME Foundation VIP Access Conference Pass to MINEXCHANGE 2027

“Check the box” and help support the SME Foundation

The Art and Science of the Flowsheet

Get your SME Foundation VIP Access Conference Pass to MINEXCHANGE 2027

“Check the box” and help support the SME Foundation

The Art and Science of the Flowsheet

To view this article you must be a member of SME/UCA.

A system dynamics simulating model of groundwater level changes and its impacts on land uplift

Incorporating operational modes into long-term open-pit mine planning under geological uncertainty

Related Articles

Rare earth elements in the global economy: Usage, recovery and the quest for supply security — A review

A systematic evaluation of localization signals in the advancing openpit mine Aitik

From foundation to future: Revisiting AI integration in mining engineering education through current perspectives of students, educators and industry

Iron ore tailings reprocessing by electrostatic separation plus two-stage froth flotation

Get your SME Foundation VIP Access Conference Pass to MINEXCHANGE 2027

“Check the box” and help support the SME Foundation

The Art and Science of the Flowsheet

Get your SME Foundation VIP Access Conference Pass to MINEXCHANGE 2027

“Check the box” and help support the SME Foundation

The Art and Science of the Flowsheet

Use Single Sign On