Copper slag as a potential source of critical elements-A case study from Tsumeb, Namibia

At a time of resource consumption, it is important to study the chemical composition of mining and metallurgical wastes to prevent the dissipative loss of metals and metalloids from the mining value chain. In particular, the recovery of critical elements from wastes is an option to increase the resources of such materials that are economically significant and have an overall supply risk. In this paper we report on the chemical composition, in particular the critical element content, of granulated slag originating from historical smelting activities in the Tsumeb area, Namibia. Laboratory-based inductively coupled plasma–mass spectrometry (ICP-MS) and X-ray fluorescence (XRF) analyses as well as portable X-ray fluorescence (pXRF) demonstrate that the slags are on average enriched in base metals (Cu 0.7 wt%, Pb 2.7 wt%, Zn 4.7 wt%), trace metals and metalloids (Cd approx. 50 mg/kg, Mo approx. 910 mg/kg) as well as critical elements (As approx. 6300 mg/kg, Bi approx. 3 mg/kg, Co approx. 200 mg/kg, Ga approx. 100 mg/kg, In approx. 9 mg/kg, Sb approx. 470 mg/kg). While metals and metalloids such as As, Mo and Pb can be determined reliably using pXRF instruments, the technique has inherent limitations in evaluating the contents of certain critical elements (Ga, Sb). However, there are positive correlations between the As, Mo, and Pb contents determined by pXRF and the Ga and Sb contents obtained through ICP-MS and XRF. Thus, quantitative pXRF analysis for As, Mo, and Pb allows calculation of Ga and Sb abundances in the slags. This work demonstrates that pXRF analysers are a valuable tool to screen smelting slags for their chemical composition and to predict the likely contents of critical elements.

The purpose of this study is to appraise the use of combined ICP-MS, AAS, XRF, and pXRF data to establish the bulk chemistry and the presence of critical elements in granulated copper slags from the Tsumeb smelter, Namibia. The pXRF method is compared with ICP-MS and XRF techniques and the application of pXRF analysis for quantitative analysis of slag samples is validated. Hence, this case study contributes to our understanding of critical elements in slags and demonstrates that pXRF is a useful addition for the chemical characterization of pyrometallurgical wastes.

Smelting site
The Tsumeb smelting complex, currently operated by Dundee Precious Metals Ltd., is located approximately 430 km north of Windhoek, the capital city of Namibia, on the northern slopes of the Otavi Mountains ( Figure 1). Smelting in the area goes back to 1907, when the German Otavi Minen-und Eisenbahn-Gesellschaft constructed two lead-copper blast furnaces to smelt local dolomitic ore . The present smelter complex was constructed in the early 1960s by Tsumeb Corporation Limited to process largely sulphidic Cu-Pb ore from the Tsumeb mine, and later also ores from Kombat (Kramer and Hultman, 1973) and Khusib Springs (Melcher, 2003). The site originally featured an integrated copper and lead section (with refinery) and smaller plants that intermittently produced cadmium metal, arsenic trioxide, and germanium (Acid Plant Database, 2020;Dundee Precious Metals Tsumeb, 2020;Mapani et al., 2014, and references therein). During the early 1980s, a slag mill was built to reprocess old copper reverbatory slags, which were milled and subsequently treated by flotation. The resulting concentrate was then processed in the smelters to extract Cu and Pb. In 1986, sodium antimonite was produced (Mapani et al., 2014, and references therein). Between 1980 and 1996, the reverbatory furnaces used fuel oil, and the smelting charge was pelletized with pulverized coal, as well as quartz, chert, and lime as fluxes Mapani et al., 2014). Since the flooding of the De Wet Shaft and the subsequent closure of the Tsumeb mine in 1996 (Bowell and Mocke, 2018), copper ores from the Democratic Republic of Congo, Zambia, Mauritania, Botswana, Greece, Russia, and in particular from Bulgaria and South America have been processed in the refurbished Ausmelt furnace, with various fluxes and fuelled with locally produced charcoal and heavy furnace oil Kabbash and Smith, 2016;Mapani et al., 2014). The granulated slag is processed in the slag milling plant and the by-product sulphuric acid is sold to industrial consumers in Namibia (Dundee Precious Metals, 2020). Tsumeb blister copper production benefits distinctly from the considerable tolerance of the Tsumeb smelter towards arsenic and lead in copper concentrates feed . Resulting from the long processing history, three different types of slags can be distinguished . Granulated slag from the reprocessing of older slags (slag type III;  is the subject of this study.

Sampling
A large slag dump (approx. 40 000 m 2 ) is located adjacent to the smelting complex and contains approximately 1.5 Mt of slag. At present, the reprocessed slags comprise granulated fragments ranging in grain size from a few millimetres in diameter to predominantly powder-size due to granulation and/ or milling Figures 1, 2, 3). During sampling in 2018, the slag heap was divided into 20 equal sectors (approximately 1 000 m 2 ). Each sector was covered by a number of parallel traverses to obtain a representative 5 kg sample of powdered to millimetre-sized slag pieces from each sector (sample numbers TSS1-TSS20). An additional slag hand Copper slag as a potential source of critical elements -A case study from Tsumeb, Namibia    (Table I).

Sample processing and laboratory-based analysis
XRD analyses were performed on pulps (<2 µm), using an X'Pert Pro (PANalytical) instrument with data collector and an X'Pert HighScore system equipped with a Cu-LFF (Empyrian) tube and an ADS tool at the Institute of Disposal Research (IDR) at Clausthal University of Technology (TUC). Qualitative evaluation was done using the X'Pert HighScore software from PANalytics ( Figure 4).
Electron microprobe (EMP) mapping of As, Cu, Ni, Pb, and Sb was done on a slag hand-specimen mounted in epoxy resin at the IDR, TUC (Figure 3). An In-As alloy was used for As (Lα1), chalcopyrite for Cu (Kα), pentlandite for Ni (Kα), crocoite for Pb (Mα), and stibnite for Sb (Lα) calibration. EMP analyses were performed using an acceleration voltage of 15 kV and a 20 nA beam current.

Portable X-ray fluorescence spectroscopy
Chemplex sample cups were filled with milled sample powders. Prolene TM thin films were used to guarantee simple, and at the same time comparable, analytical settings. All sample cups were    backfilled with stuffing fibre. Analysis was carried out using a Niton XL3t 900 hand-held XRF instrument connected to a radiation protection chamber at the IDR (TUC). The analyser is equipped with a 50 kV Ag target X-ray tube. Analyses were done in 'environmental mode -minerals with Cu/Zn', with a total measurement time of 100 seconds (Table II). Analyses were repeated five times. To test reproducibility and homogeneity two sample cups were prepared for each pulp; the results were almost identical. The instrument was calibrated using the following certified reference materials ( (Table III). Precision (i.e. the degree to which repeated measurements under unchanged conditions show the same result) and accuracy (i.e. proximity of measurement results to the true value) are important parameters to evaluate the quality of chemical analyses. Precision of pXRF values can be assessed via the percentage relative standard deviation (RSD), and accuracy via the relative difference (%RD), using the criteria after Jenner (1996) and Piercy and Devine (2014). Most major elements, metals, and metalloids show on average excellent (RSD 0-3%) to good (RSD 7-10%) precision, with the exception of MgO, Cd, and Rb (RSD >10%). Considering average values, As, Ba, Cu, Mo, Pb, Sn, and Zn show excellent precision (RSD 0-3%) and Sb very good precision (RSD 7-10%).
The average accuracy of pXRF data, compared to ICP-MS data and XRF data, is variable. Elements/oxides like SiO 2 , TiO 2 , Fe 2 O 3 , Cd, Cu, Mo, Pb, Sr, and Zn show an acceptable accuracy (%RD ≤20) compared to the ICP-MS and XRF values. An acceptable accuracy is also evident for MnO, CaO, Ba, and Sb compared to the XRF data, but accuracy is only within the 23-36%RD range compared to ICP-MS results. The average accuracy of As is close to 20%RD. Accuracy for Al 2 O 3 is poor, as well as for MgO, Rb, and in particular Sn.

Slag mineralogy
Granulated slags from Tsumeb are largely composed of X-ray amorphous substances, in particular vitreous materials, which cannot be distinguished by XRD. This feature is clearly reflected in all XRD patterns by the lack of clearly defined X-ray peaks ( Figure 4). Only (synthetic) augite and quartz could be identified in most slag samples. In addition, an uncommon spinel with Cu-Zn-Al compounds is present, as well as unknown Mg-Al oxide, Mg-Fe oxide, and Pb oxide phases. The 'quartz' proportion can The Journal of the Southern African Institute of Mining and Metallurgy VOLUME 121 MARCH 2021 probably be ascribed to the high-temperature, low-pressure SiO 2 compound tridymite and/or cristobalite ( Figure 5). No sulphide minerals, intermetallic compounds, pure metals, or arsenates (cf. ) could be detected in the sampled materials. Element mapping of As, Cu, Ni, Pb, and Sb (sample TSSF) reveals a largely very heterogeneous distribution of these elements within granulated slag powder (Figure 3) as well as within slag particles. Moreover, imaging gives clear evidence for predominance of glassy fragments in quenched milled slag. Sulphides and oxides are rare. There is a certain match between element distribution maps of As and Sb, as well as of Cu and Pb, but the distinctly greater abundance of As over Sb and Pb over Cu is obvious.

Prediction of geochemical composition from pXRF data
XRF and ICP-MS analysis are well-established techniques for determining the composition of geological materials and metallurgical wastes (e.g. Potysz et al., 2015). Such instruments used for elemental analysis must be operated in a controlled laboratory environment. Progress in XRF instrumentation has opened the way to measurement of materials in the field by means of portable XRF analysis. Portable XRF can screen for multiple elements simultaneously in a variety of materials, using one device with minimal sample preparation. However, many critical elements are only poorly determined in materials by the energy-dispersive pXRF technology (Gallhofer and Lottermoser 2018). Comparison of laboratory-generated XRF and ICP-MS data with pXRF data reveals specific element trends and may allow predictions of likely chemical compositions (Figure 7).
The chemical composition of slag samples analysed by pXRF is reported in Table II. To compare pXRF with ICP-MS/AAS and XRF data, linear regression functions were calculated for pXRF -ICP-MS/AAS and pXRF -XRF data-sets, which reveal very good correlations (R² ≥ 0.90) for As, Ba, Cd, Mo, P, Pb, Si, Sn, Sr, and Zn ( Figure 7). Good correlations (R² ≥ 0.80) are shown by Ca and Mn. Cu shows distinctly better correlations for the pXRF -ICP-Copper slag as a potential source of critical elements -A case study from Tsumeb, Namibia  Figure 7). Correlation of ICP-MS and XRF data is good for most elements (R² ≥ 0.80), and for major elements and Sb within an acceptable range (R² ≥ 0.70). Analysis of critical elements using pXRF is notoriously difficult. In the studied slags, pXRF provides precise and accurate data for only As. Other critical elements (e.g. Ga, Sb, Sn) can be detected; however, accuracy is either poor or concentrations are close to the detection limit. Cobalt is difficult to analyse by pXRF, in particular if Fe is present (Sieber and Pella, 1986), due to X-ray peak overlap (Gallhofer and Lottermoser, 2018), although concentrations are in the 100-200 mg/kg range.
Calculation of correlation coefficients for element pairs from the ICP-MS and pXRF data-sets reveals clear positive correlations of Ga, Sb, and Sn (analysed by ICP-MS) with As (analysed by pXRF), as well as a less certain correlation of Mo and As. Current commercial pXRF detectors are not capable of analysing Ga routinely. Only a few detectors can analyse for Ga if concentrations are well above 100 mg/kg (Williams-Thorpe, 2008), and if the sample matrix allows Ga detection (Lemière, 2018). For the Tsumeb slags, considering that Ga concentrations correlate with the As and Mo concentrations, Ga contents can be indirectly determined by pXRF analysis of As and Mo. pXRF analyses of both As and Mo are of excellent precision, and accuracy is acceptable for As (RSD = 1.79; %RD ICP-MS = 22; %RD XRF = 25; averages) for both pXRF -ICP-MS and pXRF -XRF data-sets and is good for Mo (RSD = 1.28; RD ICP-MS = 7.39; %RD XRF = 4.47; averages; Figure 8). Hence, the following linear regressions allow calculation of Ga concentrations: Antimony is another critical element enriched in the Tsumeb slag, although a certain portion has already been extracted. As Copper slag as a potential source of critical elements -A case study from Tsumeb, Namibia The Journal of the Southern African Institute of Mining and Metallurgy direct analysis by pXRF of Sb is acceptable, but not of good precision (RSD = 3.53; %RD ICP-MS = 23; %RD XRF = 10), indirect verification of pXRF-derived Sb data is recommended using Pb pXRF-derived data (Bero et al., 1993). The following linear formula allows the indirect calculation of Sb contents using Pb values: [3] In addition, Sb contents can be verified from As: Thus, a direct analysis of the critical element As enriched in Tsumeb granulated slag using low-cost on-site pXRF analysis is possible. In addition, indirect analysis of Ga and Sb is feasible, using As and Mo or Pb and As pXRF-derived data, respectively.

Reprocessing potential of Tsumeb slags
Slag compositions are influenced by the metallurgical process used and the composition of smelted ores and fluxes. At the Tsumeb smelter, ores of the Tsumeb deposit as well as other deposits close by in the Otavi Mountain Land were originally processed, which are well-known for their metal and metalloid enrichment (i.e. Pb, Cu, Zn, Ag, As, Sb, Cd, Co, Ge, Ga, Au, Fe, Hg, Mo, Ni, Sn, W and V) (Bowell, 2014;Frimmel, Deane and Chadwick, 1996;Melcher, 2003;Melcher, Oberthür and Rammlmair, 2006;Pirajno and Joubert, 1993). Overall, approximately 30 Mt ore at 10% Pb, 4.3% Cu, and 3.5% Zn plus significant quantities of As, Sb, Ag, Cd, and Au were extracted from the Tsumeb deposit between 1907 and 1996 (Melcher 2003). Over time, these ores and other base metal ores from mines close by, and since 1996 from several deposits overseas, were processed in the Tsumeb smelter complex, producing Pb metal, blister Cu, and Zn concentrates with by-products Ag, Au, Ge, and Cd, and intermittently As and Sb (Kramer and Hultman, 1973;Kamona et al., 1999;Mapani et al., 2014). Therefore, the studied slags do not arise from the processing and smelting of a single ore. The chemical composition of the studied slags is the result of the metallurgical treatment of diverse base metal ores and reprocessing of older slag wastes. The Tsumeb smelter is one of few smelters worldwide that is capable of processing mixed Cu-Pb ore with high As contents , and Tsumeb sulphidic ore is known for its arsenate content (Bowell, 2014). Thus, the very high As concentration detected within the metallurgical waste is a logical consequence; although some arsenic was temporarily extracted for production of arsenic trioxide (cf. Acid Plant Database, 2020). In the past, poor recovery technologies for base metals, trace metals, and metalloids as well as critical elements led to the observed element enrichments.
Tsumeb slags are enriched in the critical elements As, Bi, Ga, In, Sb, Sn, Te, and U. To date, the Southern African mineral processing and smelting industry has extracted Cu, Pb, Cd, and Sb from Tsumeb slags and flotation tailings Svoboda, Guest, and Venter, 1988). In addition, the recovery of Co from slags is possible, as demonstrated by Jones et al. (2002). Moreover, there are numerous studies that focus on recovery of trace elements from mining and processing residues using different technical approaches to recycle high-tech elements (e.g. Anand, Kanta Rao, and Jena, 1980;Gbor, Ahme, and Jia, 2000;Gbor, Hoque, and Jia, 2006;Tshiongo, Mbaya, and Maweja, 2011;Tümen and Bailey, 1990;Yang et al., 2010).
The Tsumeb copper slag is an under-exploited residue, from which metals, metalloids, and critical elements may be derived, e.g. Sn . Successful recovery of raw materials from the existing waste would require reprocessing and adjustments to the existing metallurgical processes that would allow critical element extraction. The additional benefit of such waste valorization practices would be the prevention of metalliferous drainage and dust dispersion from the slag dump in the long term (Dundee Precious Metals Tsumeb, 2020; Kříbek et al., 2014Kříbek et al., , 2018Mapani et al., 2014). For example, the amorphous glassy slag matrix exhibits a certain susceptibility to leaching during weathering (Mostafa et al., 2001;Tshiongo, Mbaya, and Maweja, 2011). Thus, recovery of additional metals from nonferrous slags as by-products would result in two benefits: it could yield critical raw materials and prevent the development of significant environmental impacts from waste disposal.

Conclusion
This study aimed to chemically characterize the granulated copper slag at the Tsumeb smelter site and to demonstrate its potential as source of critical raw materials. In addition, it was shown that pXRF is a useful addition to the techniques for chemical characterization of pyrometallurgical wastes. The results of this study demonstrate that (i) Tsumeb slags are enriched in base metals (Cu, Pb, Zn), trace metals and metalloids (Ba, Cd, Mo, S, Se), and critical elements (As, Bi, Ga, In, Sb, Sn, Te, U) (ii) The critical element As can be determined in smelting slags by pXRF at excellent precision and accuracy (iii) Other critical elements like Ga and Sb may be determined using element proxies (As, Mo, Pb) and simple linear regression functions (iv) pXRF can be used as an additional low-cost tool for screening the chemical composition of smelting slags.