**5. Discussion**

## *5.1. Classification Systematics*

The classification of the mineralized rocks of the Khalzan Buregtei deposit applying traditional methods of magmatic petrology (QAPF and TAS, see Tables 1 and 2) leads to obviously inappropriate results due to a lack of consistency across the different classification schemes. Furthermore, the classification of some samples fails to unequivocally define rock types, as compositions do not match the compositional ranges for which the schemes were defined (R1-R2, see Tables 1 and 2). This mismatch is thought (see Section 4.2) to be caused by chemical variations due to post-magmatic mineral formation. In particular, the TAS classification for plutonic rocks, derived by [22] from the classic TAS scheme for volcanic rocks, is strongly based on Na2O and SiO2 concentrations and thus notably responds to post-magmatic processes. It is therefore concluded not to use the TAS or QAPF schemes for highly mineralized alkaline granitoid rocks such as those in the centre of the Khalzan Buregtei deposit. This is in accordance with [6], who proposed a more simple classification for the rock suite of the Khalzan Buregtei massif into three classes of barren and ore bearing metasomatites and unaltered rocks of the main intrusion phase, see Figure 1.

The multication parameters R1 and R2 combine concentration data of Si, Ti, Al, Fe, Ca, Mg, K, as well as Na and thus provide the opportunity to display the chemical composition of rocks and individual rock-forming minerals (Figure 8). Consequently, the effects of post-magmatic mineral reactions altering the intrusive alkaline magmatic rocks can be graphically visualized by these parameters. Figure 8 displays the R1-R2 plot of unaltered (blue mark) and metasomatic (red mark) rocks of the Khalzan Buregtei massif. The latter samples show significantly higher R1 values in contrast to the unaltered ones and show an elevated abundance of zircon within the ore cluster. The evolution towards the point representing quartz along the line connecting feldspars and quartz suggests silicification as one major alteration feature. Thus, the formation of post-magmatic quartz can be linked to the increase of the R1 parameter parallel to the pink vectors in Figure 8. Albitization shifts rock compositions towards the albite point. However, this effect is less pronounced in the diagram due to the proximity of the points representing albite and K-feldspar. The formation of other post-magmatic Ca-bearing minerals like fluorite, REE-carbonates, and gittinsite is reflected by increasing R2 values as indicated by the vertical shift along the purple vectors in Figure 8, which can be particularly well observed for the KB G samples. The influence of Mg on the R2 value can be neglected due to the very low concentration of this element in the entire samples series. Due to the R1 and R2 value of arfvedsonite and aegirine, decomposition of these minerals enhances the effect of the elevated R1 values described above.

**Figure 8.** R1-R2 multication plot of the KB G and S samples, classification after [21] indicated by stippled grey lines. Fluorite (Flu) and quartz (Qtz) compositions plot outside the diagram and are indicated by their R1 and R2 value at the hosting axis. The blue mark indicates unaltered rocks of the Khalzan Buregtei Massif according to [5] (Syenites [5]) and Gronen (unpublished) (Syenite KB). Red mark highlights ore bearing metasomatic rocks of the KB S and KB G dataset. Purple vectors indicate the change of R2 value of samples containing CaO > 1.8 wt. %. Mineral abbreviations: Zrn = zircon, Gtt = gittinsite, Alb = albite, An50 = anorthite, Or = orthoclase K-feldspar, Arf = arfvedsonite, Aeg = aegirine. Composition of Arf and Aeg according to [24].

While traditional magmatic rock classification turns out to be inappropriate, which is also displayed in Figure 8 showing rocks with similar chemical and mineralogical compositions that plot into different fields, the pair of numeric R1-R2 values is nevertheless indicative of the petrographic character of each sample. Being based on whole rock major element data, these numeric values are appropriate for characterizing such complex metasomatically-altered rocks and therefore are considered as highly effective and unambiguous quantitative tools particularly suitable for distinction and visualization of altered or mineralized alkaline rocks and their rare metal ore bodies. Multication parameters are additive numeric values and consequently are a valuable base for modelling and geostatistics [25]. As whole rock chemical data are abundantly available already at pre-feasibility levels, multication parameters are also recommended for use in 3D modelling during exploration work or mine planning. While they fully characterize compositional rock properties, multication parameters are independent of the subjective impression or degree of skill of an individual petrographic investigator. Thus, they represent an important step to deposit characterization by quantitative parameters, which also is a prerequisite for any geometallurgical evaluation.

The geometallurgical approach considers mineralogical features of deposits in a quantitative manner with respect to their significance for mine planning and mineral processing [25]. The main focus lies on the most effective exploitation by accurate adjustment of the entire process chain from mining, mineral processing, and metallurgical procedures to mineralogical and textural ore properties [25–27].

#### *5.2. Textural Properties of Ore Minerals and Their Use for Optimized Mineral Processing*

Textural properties are crucial for the technology and costs required for ore processing. Therefore, identification and quantification of textural characteristics are as essential for ore evaluation as determination of ore mineral grade. Most ore minerals observed in the centre of the Khalzan Buregtei deposit crystallized in post-magmatic alteration reactions concomitant with the development of textures characterized by fine grain size, partial replacement, and intricate intergrowth between ore minerals

and gangue (see Section 4.1, Figures 4 and 6). These textural properties significantly affect the ore mineral liberation size, indicating the coarsest size fraction produced in comminution that guarantees the maximum detachment of ore minerals and associated gangue [28]. For the ore-bearing rock samples of KB G and KB S at Khalzan Buregtei, the liberation size of the ore minerals was obtained by image analyses applied on the phase maps generated with the QEMSCAN© technique. The liberation size distribution of the bulk group of ore minerals comprising Zr-silicates and REE-carbonates are displayed in Figure 9A. The graphs for representative samples KB G 5, 8, and KB S 8 show that 80 wt. % of all ore minerals would be liberated in fractions varying between P80 < 45 μm and P80 < 85 μm. This is well in accordance with mineral processing test work on rare metal ores from Khalzan Buregtei conducted by [29]. Likewise, [30] suggested a comminution of <80 μm followed by scavenger flotation for REE-carbonate minerals for processing of a similar rare metal ore.

**Figure 9.** Particle size distributions of ore minerals (**A**) and ore clusters (**B**). The ore mineral groups Zr-silicates and REE-carbonates were combined into the class of ore minerals because these minerals are mostly observed in cluster assemblage. As can be obtained from the particle size distribution, the ore mineral size at P80 is significantly smaller (45–85 μm) in comparison to the particle size of the ore clusters (400–500 μm).

In general, comminution to low grain size is an energy and material-consuming step, belonging to the most expensive positions in processing operations [28,31]. In addition, although required for liberation of ore phases, low grain sizes also hamper processing activities, mainly leaving flotation as a suitable technique [28]. This enhances consumption of water and supplies (e.g., flotation chemicals) as well as of operation costs [28]. While flotation turned out to be e ffective for REE-carbonates, it cannot concentrate ore minerals with distinctly di fferent physical properties, such as zircon or pyrochlore in the same step (e.g., [28,30]). Owing to the prices of Zr, Nb, and the HREE, the latter ore minerals, which remain in the flotation tailings together with a large mass of gangue minerals, would represent a higher value than the LREE predominantly extracted from carbonates. High process costs on one hand and limited concentration e fficiency on the other hand would represent an obstacle for mining operations at the Khalzan Buregtei deposit as well as for other rare metal alkaline granites.

However, the aggregation of the ore minerals zircon, gittinsite, and REE-carbonate together with other post-magmatic phases, such as fluorite and hematite in clusters characteristic of the mineralized rocks of sampling sites KB G and KB S, allows an alternative approach to more e fficient processing. Image analysis yields quantitative information on the grain size distribution of each individual ore mineral on the one hand as well as of the ore mineral clusters on the other hand. Figure 9B shows that diameters of 80 % of all ore clusters in representative samples KB G 5, 8, and KB S 8 are ≤400 to 500 μm. Consequently, liberation of ore clusters could already be achieved in size fractions, which are significantly coarser than those required for liberation of individual ore minerals (Figure 9A). It is worth mentioning that the increase in liberation size by a factor of, in part, >10 relative to the individual ore minerals, reflects the di fference between magmatic and post-magmatic mineral grain sizes. The clusters, which predominantly formed as post-magmatic pseudomorphs after arfvedsonite and elpidite (see Section 4.1) mainly have the size of their former magmatic precursor.

The increased particle size of the ore clusters is a textural feature useable for optimization of mineral processing because the liberation of the ore clusters needs less intensive comminution, which is connected to a decrease in energy consumption (e.g., [31]). In a process chain, the separation of barren and ore bearing material needs to take place as early as possible to enhance the overall process effectiveness due to mass flow reductions and increased yield of target mineral phases. Extracting a pre-concentrate of ore mineral clusters from the coarse particle size fraction will achieve both goals.

For the extraction of ore mineral clusters, physical properties are needed, which distinguish the target phases from the barren material. One possible property is the increased density of the ore-bearing clusters, which is due to the locally-increased modal abundance of mineralogical phases of high specific gravity in contrast to the barren feldspar-quartz dominated rock domains. This makes particles amenable to density separation. Further, the close associations of clustered ore minerals with hematite (Figure 4A,B) represents a marked di fference to the barren domains because the magnetic susceptibility of this Fe-bearing phase makes the ore clusters also amenable to magnetic separation.

Based on the detailed quantitative investigations of mineralogical and textural ore properties, a mineral-processing scheme was developed by [29]. The study demonstrates that the extraction of a pre-concentrate by dry magnetic separation on a particle fraction <250 μm was successfully conducted on a high volume ore sample taken from the Khalzan Buregtei deposit. Applying this optimized processing scheme, the mass flow was reduced by 55 wt. %, and an enrichment of zircon by a factor of 2.5 was achieved. As a consequence, following process steps like comminution of fractions rich in ore mineral clusters to liberation size of individual ore minerals and scavenger flotation, needed for the production of concentrates, can be conducted with the enhanced pre-concentrate having increased ore mineral content and decreased total mass.

This example shows that the mineralogical and textural properties "association of ore minerals to hematite" as well as "clustering of ore minerals" have a crucial impact on processing e fficiency and thus finally on the profit of a mining operation. For a geometallurgical approach towards optimization of the process chain, it is important to delineate volumes of mineralized rock that are characterized by consistent mineral processing responses such as geometallurgical domains [25,32,33]. In the case

of the Khalzan Buregtei deposit both properties are considered as key features for the definition of a geometallurgical domain.

The abundance of ore clusters is described for other alkaline massifs by several authors [34], for example, report clusters of REE rich apatite and allanite [(CaREE)(Al2Fe<sup>2</sup>+)(Si2O7)(SiO4)O(OH)] connected to Biotite-magnetite veins of the Loch Loyal syenite and conjugated REE-carbonate veins. Further, the replacement of arfvedsonite by hematite-quartz clusters is described by [35] for peralkaline arfvedsonite granites of the Amis complex in Namibia, which are connected to a high abundance of HFES silicates like zircon.

Clusters formed by gittinsite and zircon in association with secondary quartz replacing elpidite and arfvedsonite were also observed by [1] in the rare metal prospects of the alkaline granite at Strange Lake in Canada. These authors, too, describe post-magmatic minerals hematite, quartz and fluorite that formed in close association to ore minerals. The studies from different areas show that rare metal/HFSE enrichment in alkaline granitoid rocks is associated to the post-magmatic formation of ore-bearing clusters replacing precursor magmatic minerals. Consequently, application of the optimized beneficiation strategy, as outlined in this study, to those other rare metal alkaline granitoid deposits could be a step to improve economic mining operations.
