Redistribution of Minor and Trace Elements During Roasting of Cu-Rich Complex Concentrate in Inert Atmosphere

23 elements/isotopes were analyzed during the spot measurements and element-mapping - ³⁴S, ⁵⁵Mn, ⁵⁷Fe, ⁵⁹Co, ⁶⁰Ni, ⁶⁵Cu, ⁶⁶Zn, ⁶⁹Ga, ⁷⁵As, ⁸²Se, ⁹⁵Mo, ¹⁰⁷Ag, ¹¹¹Cd, ¹¹⁵In, ¹¹⁸Sn, ¹²¹Sb, ¹²⁵Te, ¹⁸²W, ¹⁹⁷Au, ²⁰²Hg, ²⁰⁵Tl, ²⁰⁶Pb, and ²⁰⁹Bi. This list was replicated from the recent work of George et al.^[35] wherein they performed LA-ICP-MS measurements on mineral assemblages involving the base metal sulphides (BMS) and complex minerals such as tetrahedrite. Similar combination of minerals existed in the complex Cu-rich Garpenberg concentrate studied in this work.

The data of LA-ICP-MS spot measurements was processed in a time-resolved mass spectrometry software package—‘iolite 4’ (© The iolite Team,^[45]) for data-reduction. The signals of count/s (cps) corresponding to the time-period between 1 sec after ‘Laser-start’ till 5 sec prior to ‘Laser-stop’ in each laser-ablation cycle were selected for further analysis. This approach of data-reduction performed over the measured cps-signals for Ag in tetrahedrite belonging to the initial concentrate is illustrated in Figure 4. Notably, the procedure of ‘background-subtraction’ for the measurement data was bypassed, as it was resulting in numerous negative values of cps-signals in the reduced data, especially for the trace elements.

The quantification of LA/ICP/MS cps/data into wt pct or ppm could not be satisfactorily accomplished. This could be related to certain known challenges in determining accurate element compositions from the LA-ICP-MS data for the sulphides, in addition to a few observations specific to this work listed below:

Considering the problems in quantification, the cps-signals from LA-ICP-MS after data-reduction were instead subjected to statistical evaluations.

The data of spot measurements performed over minerals in the initial concentrate was investigated using correlation matrices. For instance, in the case of tetrahedrite, all inter-element correlation coefficients between cps-signals for each spot were determined by generating a correlation matrix. Consistently high correlations over multiple spots between Sb (a major element in Ttr) and a specific trace element would suggest the presence of that trace element in the initial tetrahedrite. A similar such approach of evaluating occurrence relationships using inter-element correlation matrix was demonstrated by George et al.^[29]

The cps-signals from the chemical mapping of melt-nugget microstructure were investigated through cluster analysis. Clustering of LA-ICP-MS data has been used previously for classification of elements in an assemblage of different phases.^[40‐43,65] In the current article, the clustering was performed using the cps-signals from mapping of the major and minor elements—Cu, Fe, Zn, Pb, S, Ag, Sb, As, S. Prior to the cluster-analysis, the cps/signals for each of these elements were suitably standardized (using the transformation expression, (x− σ)/µ, σ = standard deviation, µ = mean). The algorithm of k-means clustering was used which divides the whole data-set into a specified number of clusters and allocates each data-point to one of those clusters. This clustering method is based on minimizing the Euclidean distance between the data-point and the mean of its assigned cluster. The clustering was executed using the Statistics and Machine Learning Toolbox of MATLAB 2020b (© 1994-2021 The MathWorks, Inc.). The aim of clustering was to replicate the different phases in the melt-nugget microstructure as these individual clusters. The intensities in cps of the trace elements in each of these clusters were subsequently analyzed for their classification in these phase-fractions in the melt-nugget.

Figure 7 illustrates the statistical method used in this work of determination of the significant trace elements present in minerals in the initial concentrate. Motivated from the concept of correlation matrix, element-pair-wise (for instance, Cu-vs-Co) values of correlation coefficients for each LA-ICP-MS spot measurement were collated into histograms to create a ‘matrix of correlation-histograms’. In case of Figure 7, the inter-element correlation behavior is all-together presented for the 47 spot measurements of Ttr. Those histograms which have the frequency bars closer to the right side of the plot signify a strong correlation between the cps-signals of the corresponding element-pair. Additionally, a color-scheme is used for better visualization—leftmost bar in blue (poor correlation), rightmost bar in red (strong correlation), with smooth color-transition of bars along the x-axis.

The histogram-matrix in Figure 7 shows obvious strong correlations among the major elements in Ttr—Cu, Ag, Fe, Zn, Sb, As, S. To identify the prominent trace elements in Ttr, the histograms corresponding to Sb (a major element in Ttr) with the elements ϵ {Mn, Co, Ni, Ga, Se, Mo, Cd, In, Sn, Te, W, Au, Hg, Tl, Bi} were closely observed. All Sb-related correlation-histograms are in the row and column highlighted with broken lines in Figure 7. Relatively high correlations of Sb are found with Mn, Cd and Tl (marked with black-edged rectangles in Figure 7). Based on these correlation trends, it is therefore estimated that Mn, Cd and Tl are present in trace concentrations in Ttr in the initial Garpenberg concentrate.

Notably, Sb additionally shows sufficiently good correlations with Pb in Figure 7. This can be related to the known close-association between Gn and various Ag-rich phases in the Garpenberg deposit, reported by Sandecki et al.^[67] Gn and Ttr were found to frequently co-exist, one occurring often in the matrix of another as micro-inclusion. The presence of Gn micro-inclusions in Ttr would justify such trends of good correlations between Sb-vs-Pb in the data of LA-ICP-MS spot measurement over Ttr.

Similar inter-element correlation-histograms are presented in Figure 8 for the other minerals—Ccp (6 spots), Sph (5 spots), Ag-Sb (1 spot) and Gn (5 spots). For identification of relevant trace elements, the correlation-histograms corresponding to Cu for Ccp, Zn for Sph, Ag for Ag-Sb and Pb for Gn were particularly analysed. Ccp did not exhibit prominence of any trace element. Whereas, the other histogram matrices in Figure 8 demonstrated that Sph has Mn, Cd, Hg; Ag-Sb has Cd, Hg; and, Gn has Tl in trace concentrations.

Figure 9 presents the cps-signals of the identified relevant trace elements—Tl, Mn, Cd and Hg for various minerals in the initial Garpenberg concentrate. It shows that Tl is relatively much higher in Gn compared to the other sulphide minerals. Similarly, Sph is the main carrier for Mn, Cd and Hg. Cd is present in comparable concentrations in Ttr and Sph.

It may be noted that Sph is the principal host of trace elements amongst various minerals in the initial concentrate. This is in line with the previous studies which highlighted the role of crystal structure of Sph (composed of FCC lattice with tetrahedrally coordinated Zn and S) in facilitating accommodation of a wide range of trace elements.^[26,68‐70] The crystal structure of Ttr is considered as a derivative of the atomic arrangements in Sph.^[71,72] Such an attribute could account for the similar intensities of one of the relevant trace elements—Cd in Sph and Ttr.

Regarding the relatively high intensities of Tl corresponding to spot measurements over Gn in Figure 9, Cook et al.^[25] also reported correlation between Tl and Pb in their LA-ICP-MS measurements over bornite. They attributed the coherent and flat time-resolved Tl- and Pb-profiles in laser data to the occurrence of homogeneously distributed Pb-Tl sulphosalts. In the current investigation, Figure 9 suggests that galena is likely the major host of Tl in the initial Garpenberg concentrate. However, determination of the actual Tl-bearing phase in galena is out of scope of this work.

Figure 11 presents the elemental composition data obtained through SEM-EDS for the phase pointed by ‘Arrow 3’ in Figure 10(a). Evidently, this phase is chemically close to bornite, Cu₅FeS₄ (theoretical composition - 50 mol. pct Cu, 10 mol. pct Fe and 40 mol. pct S).

Considering the CuS-FeS-ZnS ternary phase diagram shown in Figure 12, there exists an extensive solid-solution phase—iss^[12] at 700°C. Consequently, the initial Ccp transforms to iss-phase at high temperature through solid-state assimilation of co-existing Fe- and Zn-bearing minerals, as elaborated previously in Prasad et al.^[21]. The extent of additional Fe and Zn going in the initial Ccp is dictated by—(1) the overall elemental composition of the initial concentrate (i.e. Fe wt. pct beyond what is existing as Ccp, and Zn wt pct in the initial concentrate), (2) the compositional limit of the iss-phase set by its isotherm at 700°C shown in Figure 12. Now, based on the chemical composition of the initial Garpenberg concentrate presented in Table I, the point in the phase diagram representing the assimilation by initial Ccp of all available Fe and Zn in the initial concentrate is calculated and marked by a red-square in Figure 12. It is expected that the transformation of Ccp to iss-phase would follow the compositional change in the direction of the indicated arrow. This arrow is along the line joining the position of CuS:FeS = 50:50 (initial composition of Ccp, marked by point A) and the calculated Fe+Zn+Ccp position shown by the red-square (point B). The expected linearity in composition change along the line AB during the iss-phase evolution is due to the fact that there are essentially two minerals participating in the iss-phase evolution—Ccp and Sph. The Fe-rich Py mineral could not be detected in this concentrate during the QEMSCAN [Figure 6(a)] and XRD^[21] measurements. The compositional change would however be restricted till point C on the 700°C-isotherm, as the red-square lies outside the iss-phase field. Point B and C are at similar Fe-levels, but the difference is significantly high in ZnS (mol. pct) between these two points. Consequently, if the compositional change stops at point C, there will be a significant amount of surplus-Zn available in the mineral assemblage, un-assimilated in the iss-phase. The actual composition of the iss-phase in the melt microstructure measured by SEM-EDS are also overlaid in Figure 12, which are mostly located near point D on the isotherm. Notably, the points—D and B are at similar Zn-levels. The final iss-composition at 700 °C being around point D instead of point C means there has been an expulsion of certain amount of Cu from the iss-phase to gain the surplus-Zn from the mineral assemblage. This would explain the occurrence of bornite (the Cu-rich phase-fraction) in the melt-nugget microstructure in Figure 10(a). It is likely that the Bor precipitated out to facilitate the total assimilation of Sph in the iss-phase at 700 °C.

In a strictly Cu-Fe-S system, Bor and iss-phase are known to form a low-melting eutectic below 800°C.^[82,83] Inclusion of additional elements such as Zn, Pb, Ag, Sb, As (as in the presently investigated system of the Garpenberg concentrate) would likely reduce the melting point of this eutectic-phase combination further down. Interestingly, Bor exists primarily in the gaps between the solidified intergrowth phases in the microstructure in Figure 10(a). The thermodynamic phase-relations of Bor with the co-existing phase-fractions (such as with iss) together with its microstructural attribute suggest that Bor was supposedly the last liquid to solidify during cooling of the roasted calcine.

In addition to the iss-PGn, PT-PGn intergrowths and Bor-phase-fraction, the microstructure in Figure 10(a) also comprises of a bright-yellow phase-fraction which is Ag-rich [Figure 10(g)]. EDS-measurements over this phase revealed that its composition is approximately 95Ag-5Sb (wt pct), which is presented in Figure 13(a) along with the EDS-measurements corresponding to the Ag-Sb mineral phase in the initial concentrate. Clearly, the newly formed phase is richer in Ag compared to the initial Ag-Sb (comprising approx. 75 wt pct Ag). Apparently, there has been mobilization of Ag from other mineral phases to join the Ag-Sb phase during roasting. Since Ttr and Ag-Sb were the principal Ag-carriers in the initial concentrate [Figure 6(d)], this significant Ag-enrichment in Ag-Sb phase during roasting would signify transfer of Ag from Ttr into Ag-Sb.

Ag is known to substitute up to 6 Cu atoms at the monovalent sites in a formula unit of Cu₁₂Sb₄S₁₃ (theoretical tetrahedrite). Also, in most of the natural tetrahedrites, Cu occupies a maximum of up to 10 (6 Cu⁺, 4 Cu²⁺) of its designated 12 sites. The balance 2 divalent sites are generally occupied by Fe²⁺ and Zn²⁺.^[84] Such Cu-Ag compositional attributes can be seen in Figure 13(b) obtained from EDS measurements over Ttr in the initial Garpenberg concentrate. Ttr exhibits wide-compositional variation with a linear relationship between Cu and Ag. There is evidently a full-range of solid-solution in Ttr constituted by two end members—(1) Ttr containing 6 Ag atoms p.f.u. (per formula unit), and (2) Ttr containing 10 Cu atoms p.f.u. During roasting, Ttr transforms into pseudo-tetrahedrite (PT) in the process of forming a low-melting eutectic with Gn.^[19‐21] It can be observed from Figure 13(b) that this Ttr→PT transformation resulted in loss of Ag from the initial Ttr. The PT melt phase-fraction has a relatively very low Ag-content, implying release of Ag from Ttr at high temperature. Supposedly, the Ag-Sb mineral phase was joined by the Ag released from Ttr resulting in the formation of 95Ag-5Sb melt phase-fraction.

Figure 14(a) presents an SEM image of the microstructural site in the melt-nugget which was characterized by LA-ICP-MS for chemical mapping. The rectangular area which was mapped is highlighted, showing the rows of laser-ablation craters. The thickness of these rows is 10 µm, which is the selected spot-size for chemical mapping. Figure 14(b) is the optical image of the same microstructural site, acquired prior to the laser-ablation. The rectangular area in this microstructure undergoing subsequent laser-ablation is again highlighted in this image.

The generated dataset of cps-signals from the LA-ICP-MS chemical mapping for the major and minor elements ϵ {Cu, Pb, Zn, Ag, As, Sb, S} was appropriately standardized. The standardized dataset was repeatedly subjected to k-means clustering analysis for 8 times, only changing the pre-assigned number of data-clusters from 1 to 8 over these runs. Figure 14(c) presents the plot of summation for within-cluster point-to-centroid Euclidean distance obtained in each run. The summation distance is found to reduce significantly until 5 clusters, beyond which the decrease is relatively small. It is suggestive of the existence of optimal 5 clusters in the standardized chemical-mapping dataset. This is consistent with the existence of 5 phase-fractions in the melt-nugget microstructure—iss, PT, PGn, Bor and 95Ag-5Sb. To further validate the clustering approach, a false-colored rastered image was produced based on the cluster numbers ϵ {1, 2, 3, 4, 5} assigned to each measurement data-point. This image presented in Figure 14(d) closely resembles the phase assemblage in the rectangular area in Figure 14(b). For instance, portions in Figure 14(d) representing Cluster 1 in white-color match consistently with the light-purple colored PGn phase-fraction in the original microstructure in Figure 14(b). [The light-purple colored phase in the melt-nugget microstructure was identified as PGn earlier in Figure 10(a)]. Such an effective reproduction of microstructure of the laser-ablated area implies that each data-cluster corresponds to one of the 5 phase-fractions in the melt-nugget microstructure.

Since the 5 identified melt phase-fractions were distinctly rich in one of the elements ϵ {Cu, Fe, Pb, Sb, Ag}, the cps-signal intensities for these elements in the LA-ICP-MS chemical-mapping dataset were compared between the 5 data-clusters. These comparisons are shown in Figure 15(a) through (e).

Evidently, Cluster 1 is Pb-rich and thereby represents PGn melt phase-fraction. Similarly, Cluster 2 being Cu-rich represents Bor; Cluster 3 being Fe-rich represents iss; Cluster 4 being Sb-rich represents PT; and, Cluster 5 being Ag-rich represents 95Ag-5Sb. These relationships between data-clusters and melt phase-fractions are also in line with the matching of clusters in the reproduced microstructure in Figure 14(d) with the phase-fractions in the original microstructure in Figure 14(b). However, it is important to note that the order-of-magnitude comparison of cps-signals for these elements {Cu, Fe, Pb, Sb, Ag} between different clusters/phase-fractions in Figure 15 is not valid in this study. The spot size of 10 µm for the LA_ICP-MS chemical mapping would not be sufficiently small enough to consistently represent only a single phase-fraction. Especially, in the regions of finer intergrowths in Figure 14, the ablated material from each spot in the rastered-scan would be a mixture of multiple phase-fractions. However, as pointed out earlier in Section III-B-2-a, the clustering method could overcome this problem of phase-mixing and could effectively classify the measured spots into representative clusters.

The relevant trace elements identified earlier in Section III-A-2-a were—Tl, Mn, Cd and Hg. The cluster-wise comparison of their cps-signals (from LA-ICP-MS chemical mapping) are presented in Figure 15 (f) through (i). It can be noticed in Figure 15(f) through (g) that the intensities for Tl and Mn are relatively high for Cluster 2 (Bor). It suggests that Tl and Mn segregated in the Bor-phase-fraction during solidification of the melt. Similarly, Cd-intensities being high for Cluster 4 (PT) in Figure 15(h) would indicate the segregation of Cd in the PT phase-fraction. Notably, Hg-intensities are in similar ranges for all clusters [Figure 15(i)]. Apparently, Hg was homogeneously distributed in all the phase-fractions, and did not undergo any noticeable segregation during solidification.

The preferential redistribution of individual trace elements into selective melt phase-fractions could be associated with numerous factors. Minerals such as Ttr and Bor occur naturally as complex superstructures which can accommodate trace elements through solid-solutions, inclusions or structural defects.^[25,71,85] Concurrently, the order of solidification of multiple liquids could be relevant, as many trace elements could partition into the low-melting phases. For instance, the microstructural attribute of Bor in Figures 10 and 14 as a phase filling the gaps between intergrowth phases points towards its subsequence in the solidification-order. However, detailed explanation of the phase-preferences of individual trace elements during melt-nugget solidification necessitates further research, which is not in the scope of the present study.